Previous studies have investigated the characteristics of machine translation(MT)-mediated communication in lab settings and suggested various ways to improve it [1]. Unfortunately, we still lack an understanding of how MT is used in real-world settings, particularly when people use it to support face-to-face communication. In this paper, we report on a field study of a multilingual workshop where children from various language regions used MT to communicate with each other. We investigate how children use various information such as nonverbal cues and drawings to compensate for the mistranslations of MT. For example, children tried to understand the mistranslated messages by reading alternative translations and used web browsers to search for pictures of unknown objects. Such findings provide insights for designing future multilingual support systems.

Various language services exist to support the listening comprehension of non-native speakers (NNSs). One important service is to provide NNSs with real-time transcripts generated by automatic speech recognition (ASR) technologies. The goal of our research is to explore the effects of ASR transcripts on the listening comprehension of NNSs and consider how to support NNSs with ASR transcripts more effectively. To reach our goal, we ran three studies. The first study investigates the comprehension problems faced by NNSs, and the second study examines how ASR transcripts impact their listening comprehension, e.g., what types of comprehension problems could and could not be solved by reading ASR transcripts. Finally, the third study explores the potential of using eye-tracking data to detect their comprehension problems. Our data analysis identified thirteen types of listening comprehension problems. ASR transcripts helped the NNSs solve certain problems, e.g., “failed to recognize words they know.” However, the transcripts did not solve problems such as “lack of vocabulary,” and indeed NNS burden was increased. Results also show that from eye-tracking data we can make reasonably accurate predictions (83.8%) about the types of problems encountered by NNSs. Our findings provide insight into ways of designing real-time adaptive support systems for NNSs.

Many language resources are being shared as web services to process data on the Internet. As dataset size keeps growing, language services are experiencing more big data problems, such as the storage and processing overheads caused by the huge amounts of multilingual texts. Parallel execution and cloud technologies are the keys to making service invocation practical. In the Service-Oriented Architecture approach, service providers typically employ policies to limit parallel execution of the services based on arbitrary decisions. In order to attain optimal performance, users need to adapt to the services policies. A composite service is a combination of several atomic services provided by various providers. To use parallel execution for greater composite service efficiency, the degree of parallelism (DOP) of the composite services need to be optimized by considering the policies of all atomic services. We propose a model that embeds service policies into formulae and permits composite service performance to be calculated. From the calculation results, we can predict the optimal DOP for the composite service that allows the best performance to be attained. Extensive experiments are conducted on real-world translation services. The analysis results show that our proposed model has good prediction accuracy in identifying optimal DOPs for composite services.

One of the main features of The Language Grid is its support for service composition, i.e. creating new language services that meet user requirements by combining the existing ones. Despite the potential of service composition, such a service-oriented computing (SOC) application may experience cascading failure when a disruption on one or more component services is propagated to the composite services that combine them. As the number of language services grows, composite language services will become more common, and thus understanding cascading failure among language services becomes more important. This chapter investigates how failure may propagate among language services and how to improve language service tolerance to cascading failure. To this end, the dependency between language services is modeled as service network on which cascading failure is simulated and analyzed. We also generated service networks in scale-free, exponential, and random topology to analyze how cascading failure occurs in different topology. The simulation reveals that service networks with scale-free topology have better cascading-failure tolerance compares to that of other topology.

Transferring knowledge to other people in different languages is difficult because of gaps in languages and cultures. It makes the knowledge transfer more difficult when the recipient is young, because the comprehension and language ability of the young are incomplete. To better understand and design language services, this chapter introduces a communication protocol that meets requirements of agriculture support in rural areas, and fully delineates the communication environment by elucidating the field issues comprehensively
solutions are considered. The field experiment conducted involves agriculture support in Vietnam. In the context of agriculture support in rural areas, there exist several issues such as the requirement of timely knowledge transfer with high translation quality, and multilingual communication between youths and experts where gaps in language ability and expertise should be considered and addressed.

Self-organization has been proposed to be implemented in complex systems which require the automation capabilities to govern itself and to adapt upon changes. Self-organizing systems can be modeled as multi-Agent systems (MAS) since they share common characteristics in that they consist of multiple autonomous systems. However, most existing MAS engineering methodologies do not fully support self-organizing systems design since they require predefined goals and agent behaviors, which is not the case in self-organizing systems. Another feature that is currently not supported for designing self-organizing MAS is the separation between the design of agent behaviors and behavior adaptation, i.e. how agents adapt their behaviors to respond upon changes. To tackle these issues, this paper proposes a role modeling method, in which agent behaviors are represented as roles, to design how agents perform behavior adaptation at runtime by switching between roles. The applicability of the proposed role modeling method is evaluated in a case study of a self-organizing smart transportation system.

A design education program must unite traditionally siloed disciplines, spanning rigid academic boundaries from engineering to social sciences. Work to establish the Kyoto University Design School had to meet these challenges in promoting interdisciplinary cooperation and provides a solid model for others to build on.

The future Internet will be populated with a massive number of cooperating services due to the rapid growth of publicly available services and the adoption of service-oriented computing (SOC) into the Internet of Things. The adoption of SOC enables combining the functionalities of smart devices as combining services by means of service composition. These cooperating services form a large-scale service network where the nodes and the links represent services and the dependency between services, respectively. The dependency between services potentially causes cascading failure, where the failure of a service propagates to its dependent services. Due to the lack of research in this type of cascading failure, we analyzed cascading failure in service networks for different topology and different degree of service interdependency. We found that the number of cascading failure is somewhat linear to the average number of required services, and decays exponentially over the average number of alternate services. The latter suggests that cascading failure tolerance can be significantly improved by adding few alternate services to each required service if the average number of alternate services is currently low. In addition, we also found that scale-free topology provides better tolerance, subsequently followed by exponential and random topology.

In human communication, nonverbal information such as gestures and facial expressions often plays a greater role than language, and an increasing number of devices are designed to be intuitively controlled by gestures. However, there are some disadvantages of this intuitive interaction. One of the chief problems is that these devices have difficulty in distinguishing between unconscious and intentional gestures
they tend to respond erroneously to unconscious movements. In this chapter, authors propose a new gesture analysis method based on the dynamic model. They focused on the "exaggerated gestures" that are effectively used in, such as Japanese Kabuki, effectively used in Disney's animation, and tried to identify their common features and effects. They noted the "preparation" or "follow-through" motions just before and after the emphasized actions and each behavior can be quantified by the undershoot and overshoot value of changes in torque. These methods can provide important knowledge for analyzing features and distinguishing intentions when interacting with gestures.

Multiple official languages within a country along with languages common with other countries demand content consistency in both shared and unshared languages during information sharing. However, inconsistency due to conflict in content shared and content updates not propagated in languages between countries poses a problem. Towards addressing inconsistency, this research qualitatively studied traits for information sharing among countries inside global brands as depicted by content shared in their country-specific websites. First, inconsistency in content shared is illustrated among websites highlighting the problem in information sharing among countries. Second, content propagation among countries that vary in scales and coupling for specific content categories are revealed. Scales suggested that corporate and customer support related information tend to be shared globally and locally respectively while product related information is both locally and regionally suitable for sharing. Higher occurrences of propagation when sharing corporate related information also showed tendency for high coupling between websites suggesting the suitability for rigid consistency policy compared to other categories. This study also proposed a simplistic approach with pattern of sharing to enable consistent information sharing.

Many language resources have been shared as web services to process data on the internet. As data sets keep growing, language services are experiencing more big data problems, such as challenging demands on storage and processing caused by very large data sets such as huge amounts of multilingual texts. Handling big data volumes like this requires parallel computing architectures. Parallel execution is one way to improve performance of language services when processing huge amounts of data. The large data set is partitioned and multiples processes of the language service are executed concurrently. However, due to limitation of computing resources, service providers employ policies to limit number of concurrent processes that their services could serve. In an advanced language application, several language services, provided by different providers with different policies, are combined in a composite service to handle complex tasks. If parallel execution is used for greater efficiency of a language application we need to optimize the parallel configuration by working with the language service policies of all participating providers. We propose a model that considers the atomic language service policies when predicting composite service performance. Based on this model, we design a mechanism that adapts parallel execution setting of a composite service to atomic services’ policies in order to attain optimal performance for the language application.

Cloud computing and high performance computing enable service providers to support parallel execution of provided services. Consider a client who invokes a web service to process a large dataset. The input data is split into independent partitions and multiple partitions are sent to the service concurrently. A typical customer would expect the service speedup to be directly proportional to the number of concurrent requests (or the degree of parallelism - DOP). However, we obtained that the achieved speedup is not always directly proportional to the DOP. This may because service providers employ parallel execution policies for their services based on arbitrary decisions. The goal of this paper is to analyse the performance improvement behavior of web services under parallel execution. We introduce a model of parallel execution policy of web services with three policies: Slow-down, Restriction and Penalty policies. We conduct analyses to evaluate our model. Interestingly, the results show that our model have a good accuracy in capturing parallel execution behavior of web services.

Agent-based Web service composition has become one of the most challenging research issues. Many composition techniques and formalizations have been proposed, but they are neither mature nor flexible. They assume that each sub-task is an atomic process, hence it cannot be decomposed based on user requirements. Moreover, those techniques and formalizations are not suitable for dynamic environments such as the language service domain. Language service requires a flexible formalization to accommodate the user's language skills in conjunction with QoS. The key contributions of this paper are (i) a complete formalization that ideally reflects the nature of real applications and permits extension of the original abstract workflow (in case of failure); (ii) a novel agent-based protocol able to find satisfying solutions for this problem in real time to allow restriction and/or relaxation within the original workflow; (iii) a hybrid architecture of service-oriented computing and multi-agent systems for implementing Abstract Web service, Information analysis, and User agents. Experiments are presented to find solutions that can be executed within a feasible time and space.

This paper reports that useful search applications can be developed around "One-Touch-Character", a simplified method using digit keys originally developed for display-free touch-tone-telephones. Several input methods created in Japan have transformed into cultures characterized by their specific operations. "One-Touch-Character", one of the oldest methods, was born in 1990s and is seen as obsolete in the smartphone era. This paper introduces and compares input methods developed over the last 25 years. The drastic changes in appearance and features, from touch-tone telephones to smartphones through mobile phones, have strongly influenced input style. Though young people learnt quickly and became experts, those who could not catch the innovations became IT illiterates. To help them become positive users, we return to "One-Touch-Character". After investigating several different databases, we clarify what database characteristics suit "One-Touch-Character".

There is ongoing research on information systems that support inter-disciplinary communication for people in different fields of expertise as well as on intercultural communication for people who speak different languages. However, there are few reports of research of systems that facilitate communication between youth and experts who speak different languages. In the presence of different languages and knowledge sets, a key issue lies in the implementing communication environments becomes difficult. This paper discusses a communication protocol designed to send/receive data and knowledge between youth and experts. This paper implements intercultural communication environment is implemented by combining machine translation services with human services to enhance communication quality. Finally, we validated the communication by applying the communication environment in agriculture support in Vietnam.

The rapid growth of services and the Internet of Things vision lead to the future of Internet in which a massive number of services are available and connected to each other. In such service network, dependency between services potentially causes cascading failure, where the failure of one service can cause the failure of dependent services. Cascading failure tolerance is determined by the topology of the network and the degree of service interdependency. As to the former, we analyze cascading failure in scale-free, exponential, and random service networks. We find that scale-free topology has generally the highest tolerance. This is contrast to cascading failure in power network, where random topology provides better tolerance. For the latter, we find that the number of cascade failed nodes increases as the inverse of the average number of alternate services, e.g. functionally equivalent services. This suggests that increasing the number of alternate services can significantly improve the network tolerance if each service only has few alternate services available.

Reputation plays a crucial role in the success of e-commerce. In a commercial transaction, it is necessary to present reputation values of web services in a timely and a robust manner so as to counter the unfair ratings of malicious users. To address the time lag problem, most popular web sites use an averaging algorithm with fixed sliding windows; window size is constant and older ratings are dropped upon the arrival of new ratings. Herein, we propose a dynamic sliding window model that is capable of reflecting the reputation values yielded by the latest changes in services. Furthermore, we implement a statistical strategy to filter out unfair ratings by calculating the standard deviation of the ratings after transposing the two-dimensional linear window into the constant one-dimensional window by using linear regression. Experiments confirm the effectiveness of the proposed model, it outperforms the existing reputation system by 40% on average based on the 5 test cases examined, and also show that it can asymptotically converge to the underlying reputation value as ratings accumulate.

Multilingual knowledge sharing imposes new requirements on knowledge management systems so as to present digital knowledge resources in multiple languages. Knowledge sharing is degraded by inconsistencies such as contents omitted or altered in one of the languages. To resolve this issue, we present a mechanism for detecting inconsistencies in multilingual knowledge sharing. A state transition model is proposed to define the states of the multilingual contents,the set of actions, and the set of transition functions. Inconsistency detection rules are designed to represent the states of the multilingual contents and thus permit the identification of inconsistencies in knowledge sharing. The analysis of a multilingual Wikipedia article indicates that inconsistencies are present in multilingual contents generated by collaboration. In experiments, the proposed mechanism is applied to a test set of revision histories of multilingual articles
the outcome shows satisfactory results with an average precision of 88% in detecting inconsistencies and a recall of86%. While the proposal considers only user edit actions, it can detect inconsistencies which will be useful in allowing Natural Language Processing (NLP) based systems to synchronize multilingual contents in an early phase.

Collaborative business has been increasingly developing with the environment of globalization and advanced information technologies. In a collaboration environment with multiple organizations, participants from different organizations always have different views about modeling the overall business process due to different knowledge and cultural backgrounds. Moreover, flexible support, privacy preservation and process reuse are important issues that should be considered in business process management across organizational boundaries. This paper presents a novel approach of modeling interorganizational business process for collaboration. Our approach allows for modeling loosely coupled interorganizational business process considering different views of organizations. In the proposed model, organizations have their own local process views of modeling business process instead of sharing pre-defined global processes. During process cooperation, local process of an organization can be invisible to other organizations. Further, we propose the coordination mechanisms for different local process views to detect incompatibilities among organizations. We illustrate our proposed approach by a case study of interorganizational software development collaboration.

The availability of more and more Web services provides great varieties for users to design service processes. However, there are situations that services or service processes cannot meet users' requirements in functional QoS dimensions (e.g., translation quality in a machine translation service). In those cases, composing Web services and human tasks is expected to be a possible alternative solution. However, analysis of such practical efforts were rarely reported in previous researches, most of which focus on the technology of embedding human tasks in software environments. Therefore, this study aims at analyzing the effects of composing Web services and human activities using a case study in the domain of language service with large scale experiments. From the experiments and analysis, we find out that (I) service implementation variety can be greatly increased by composing Web services and human activities for satisfying users' QoS requirements; (2) functional QoS of a Web service can be significantly improved by inducing human activities with limited cost and execution time provided certain quality of human activities; and (3) multiple QoS attributes of a composite service are affected in different ways with different quality of human activities.

We present agent metaphor as a novel interactive system to promote the efficiency in machine translation mediated communication. Machine translation is increasingly used to support multilingual communication. In the traditional, transparent-channel way of using machine translation for the multilingual communication, translation errors are ignorable, due to the quality limitation of current machine translators. Those translation errors will break the communication and lead to miscommunication. We propose to shift the paradigm from the transparent-channel metaphor to the human-interpreter metaphor, which motives the interactions between the users and the machine translator. Following this paradigm shifting, the interpreter (agent) encourages the dialog participants to collaborate, as their interactivity will be helpful in reducing the number of translation errors, the noise of the channel. We examine the translation issues raised by multilingual communication, and analyze the impact of interactivity on the elimination of translation errors. We propose an implementation of the agent metaphor, which promotes interactivity between dialog participants and the machine translator. We design the architecture of our agent, analyze the interaction process, describe decision support and autonomous behavior, and provide an example of preparing repair strategy. We conduct an English-Chinese communication task experiment on tangram arrangement. The experiment shows that compared to the transparent-channel metaphor, our agent metaphor reduces human communication effort by 21.6%.

Proper technology transfer based on scientific data is inevitable for farmers to promote sustainable agriculture which is a common target of modern food production. In under-developing countries, illiteracy of farmers is one of the major reasons for them to receive sufficient information and knowledge for such sustainable food production. In this study, we conducted a trial to examine effectiveness of an idea to transfer agricultural knowledge and information to illiterate farmers through their children educated at school as messengers between their illiterate parents and remote experts by utilizing several ICT tools including a multilingual machine translator which can bridge the famers and even foreign experts. © 2014 IEEE.

According to the rapid spread of the Internet, the new devices and web applications using the newest multimedia technologies are proposed one after another and they become commodity in an instant. In these new web communications, the natural and intelligible interaction corresponding to the user's various demands is required. In the communication in which persons do the direct dialogue in the interaction not only on the web but also in real world, it is widely known by the psychology field that the nonverbal information which is hard to express in words such as expression of face and gesture is playing the important role. In our research, the new analysis method of interaction using the dynamical model is proposed and paid our attention to the characteristic gestures especially. These gestures are the special motions such as lively or powerful actions which used effectively in Kabuki, anime, dance and the special gestures in the speech and presentation of attracting audiences. By analyzing the mechanisms of these characteristic gestures mathematically, we can design the new interactive interfaces easily which are natural and familiar for all users.

How can we predict ICT-driven innovations that will emerge in society and daily life? Multiagent simulations can be used to predict the changes in society and daily life caused by human interaction with new technologies. Multiagent simulations have become increasingly popular as a type of micro-simulation that can represent diversity and heterogeneity of human behaviors. In contrast to traditional micro-simulations, multiagent simulations can represent individual decision-making in detail, so they can reproduce the complex phenomena that arise from the outcome of interactions between different agents. In this chapter, we introduce three cases where multiagent simulations are used to reproduce and analyze complex collective behavior
evacuation simulation, traffic simulation, and economic simulation. Furthermore, we discuss the participatory approach for realizing practical and reliable multiagent simulations.

A service invocation in a composite service is usually defined by its interface for abstraction. Once the interface is standardized and many services which have the same interface become available, a user can bind preferable ones to the composite service only by setting endpoints. Since new services become available day by day in open environment, it is crucial to automatically construct the bindings according to user's requirements. However, runtime binding construction often degrades the performance because searching a vast amount of possible bindings is time-consuming. Moreover, bindings need to be modified even during execution of a composite service when some adaptation process is applied. In this paper, we proposed a framework for constructing bindings, which is based on ATMS (Assumption-based Truth Maintenance System). The framework effectively caches possibilities of bindings and can manage changes of bindings when a runtime adaptation process is applied. We implemented a prototype of the framework and confirmed that it works fast enough for a real scale service composition. © 2012 IEEE.

The real-time interactive virtual classroom with tele-education experience is an important approach in distance learning. However, most current systems fail to meet the new challenges raised by the development of the serviceoriented architecture. First, the learning systems should be able to facilitate easier integration of increasingly dedicated services, such as language services on the Internet. Second, the learning systems must open their internal interfaces as web services to other systems, so as to enable deeper integration of these systems and easier deployment. Third, the systems are expected to provide flexible interfaces to support mobile device interaction. To address these issues, we build a prototype system, called Open Smart Classroom, by upgrading the original Smart Classroom into a service-oriented open system. With the help of Language Grid services, two Open Smart Classrooms deployed in Tsinghua University and Kyoto University are connected and experimental co-classes have been successfully held. The results of the user study show that integrating Smart Classroom and language services is an interesting and promising approach to building future multicultural distant learning systems.

The Language Grid offers language services with a standardized interface and different non-functional properties. This allows us to create a specialized composite service for our own goals simply by selecting the appropriate services. The language services are, however, provided in various formats with their own policies. In an environment for service-collective intelligence, it is essential to have many service providers join by strongly ensuring that their policies are satisfied. In doing this, we therefore we have to solve the following problems. First, service composition relies on the products of various stakeholders that belong to different organizations, such as service products and composite service designers. This makes it difficult to modify existing services in line with given requirements. Next, selection of services may impose constraints on execution. We therefore often need to apply a certain amount of runtime adaptation toward a composite service in order to enforce given policies. To solve these problems, we proposed an architecture for runtime service management called Service Supervision. Service Supervision provides meta-level execution functions for composite services. These allow operators to modify behaviors of a composite service without changing its model. Service Supervision is also capable of effectively managing a comprehensive process of runtime service selection and adaptation in order to ensure the service providers’ policies are satisfied. We implemented the Service Supervision prototype and showed that applying meta-level execution control barely decreases performance.

This chapter reports on our experiences with combining two different platforms in natural language processing research, i.e. Heart of Gold and the Language Grid, to provide more language resources available on the Web. Heart of Gold is known as middleware architecture for pipelining deep and shallow natural language processing components. The Language Grid is one of the service grid infrastructures built on top of the Internet to provide pipelined language services. Both of these frameworks provide composite language services and components. Having Heart of Gold integrated in the Language Grid environment contributes to increased interoperability among various language services. The integrated architecture also supports the combination of pipelined language services in the Language Grid and the pipelined natural language processing components in Heart of Gold to provide a better quality of language services available on the Web. Thus, language services with different characteristics can be combined based on the concept of Web service with different treatment of each combination. An evaluation is presented to show that the overhead of the integration is not significant.

In the Language Grid, automatically composing Web services is a crucial task. This task involves vertical and horizontal composition. Vertical composition consists of defining an appropriate combination of simple processes to perform a composition task. Horizontal composition consists of determining the most appropriate Web service from among a set of functionally equivalent ones for each component process. The latter is important in language services. For the horizontal composition of Web services, we propose a generic formalization of any Web service composition problem based on a constraint optimization problem (COP) and then propose an incremental user-intervention-based protocol to find the optimal composite Web service according to some predefined criteria at run-time.

When people communicate in their native languages using machine translation, they face various problems in constructing common ground. This study, based on the Language Grid framework, investigates the difficulties of constructing common ground when pairs and triads communicate using machine translation. We compare referential communication of pairs and triads under two conditions: in their shared second language (English) and in their native languages using machine translation. Consequently, to support natural referring behaviour in machine translation mediated communication between pairs, our study suggests the importance of resolving the asymmetries and inconsistencies caused by machine translations. Furthermore, to successfully build common ground among triads, it is important for addressees to be able to monitor what is going on between a speaker and other addressees. The findings serve as a basis for designing future machine translation embedded communication systems. The proposed design implications, in particular, are fed back to the Language Grid development process and incorporated into the recent Language Grid Toolbox.

Pictogram communication is successful when participants at both ends of the communication channel share a common pictogram interpretation. Not all pictograms carry a universal interpretation, however
the issue of ambiguous pictogram interpretation must be addressed to assist pictogram communication. To unveil the ambiguity possible in pictogram interpretation, we conduct a human subject experiment to identify culture-specific criteria employed by humans by detecting cultural differences in pictogram interpretations. Based on the findings, we propose a categorical semantic relevance measure which calculates how relevant a pictogram is to a given interpretation in terms of a given pictogram category. The proposed measure is applied to categorized pictogram interpretations to enhance pictogram retrieval performance. The WordNet, the ChaSen, and the EDR Electronic Dictionary registered to the Language Grid are utilized to merge synonymous pictogram interpretations and to categorize pictogram interpretations into super-concept categories. We show how the Language Grid can assist the cross-cultural research process.

In this chapter, we present a protocol for collaborative translation, where two non-bilingual people who use different languages collaborate to perform the task of translation using machine translation (MT) services. The key idea of this protocol is that one person, who handles the source language and knows the original sentence (source language side), evaluates the adequacy between the original sentence and the translation of the sentence made fluent by the other person, who handles the target language (target language side). In addition, by determining whether the meaning of the machine-translated sentence is understandable, it is ensured that the two non-bilingual people can do the above tasks without stopping the protocol. As a result, this protocol 1) improves MT quality
and 2) terminates successfully only when the translation result becomes adequate and fluent. An experiment shows that when the protocol terminates successfully, the quality of the translation is increased to about 83 percent for Japanese-English translation and 91 percent for Japanese-Chinese translation. We contributed to the Language Grid Project by proposing a new way to use MT services efficiently in real fields.

The Language Grid offers a broad array of language services such as dictionaries and translation, and cascading them enables people in different parts of the world to communicate with one another in their mother tongue. However, when cascading several translation services, words’ meanings often drift due to the inconsistency, asymmetry and intransitivity of word selection. In this section, we propose context-based coordination to maintain the consistency of word meanings. For this, we put forth a method to automatically generate multilingual equivalent terms based on the use of bilingual dictionaries. We generated trilingual equivalent noun terms and implemented a Japanese-to-German-and-back translation, cascading four translation services. The evaluation results showed that the generated terms can cover over 58% of all nouns. Translation quality was improved by 41% for all sentences, and the quality rating for all sentences increased by an average of 0.47 points on a five-point scale.

The multilingual Wikipedia is the largest existing collaboratively edited encyclopedia, where several translation communities are working towards translating Wikipedia articles. The different language communities are largely independent in terms of policy creation, behavior and community mechanisms. We conducted a case study on the Wikipedia community from a multilingual point of view to better understand community behavior. We also conducted a collaborative Wiki-to-Wiki translation experiment using machine translation tools provided by the Language Grid. Based on the findings of the two studies we designed and developed a multi-language discussion platform for Wikipedia translation communities. In this chapter, we discuss the results of the case study and a laboratory experiment and how the results are applied to facilitate the creation of multilingual collective intelligence in Wikipedia through a multi-language discussion platform.

Building and maintaining multilingual websites is necessary for organisations and communities trying to increase their global visibility. Such a localisation process might be done in different ways and the structure of a multilingual website should reflect it. However, not clearly understanding localisation patterns in multilingual websites increase the localisation cost. In previous studies, attempts were made to build specific models for multilingual websites but without differentiating between the different possible patterns. In this paper, we present a new formal graph based model, composed of the link structure and of the translation path between the different language versions of a multilingual website. We define some characteristic properties based on this model and derive four localisation patterns. This model was validated through the development of an experimental platform to extract patterns from three multilingual websites with different purposes. The model allowed for a better understanding of localisation patterns in these websites. © 2011 IEEE.

Nowadays Wikipedia has become useful contents on the Web. However, there are great differences among the number of the articles from language to language. Some people try to increase the numbers by the translation, where they should have a discussion (regarding the discussion about the translation itself) because there are some specific words or phrases in an article. They can make use of machine translation in order to participate in the discussion with their own language, which leads to some problems. In this paper, we present the algorithm "Meta Translation", to keep the designated segments untranslated, and to add the description into it. © 2011 IEEE.

Discussion pages in individual Wikipedia articles are a channel for communication and collaboration between Wikipedia contributors. Although discussion pages contribute to a large portion of the online encyclopedia, there have been relatively few in-depth studies conducted on the type of communication and collaboration in the multilingual Wikipedia, especially regarding translation activities. This paper reports the results on an analysis of discussion about translation in the Finnish, French and Japanese Wikipedias. The analysis results highlight the main problems in Wikipedia translation requiring interaction with the community. Unlike reported in previous works, community interaction in Wikipedia translation focuses on solving problems in source referencing, proper nouns and transliteration in articles, rather than mechanical translation of words and sentences. Based on these findings we propose future directions for supporting translation activities in Wikipedia. © 2011 IEEE.

We present the research and work carried out towards analyzing patterns in creating teaching materials from the Web and how this leads to the model of the multilingual learning resource system using Web service technologies for today's educators, to efficiently share, reuse and translate existing teaching materials into minority languages and make new ones. Copyright 2011 ACM.

In this demonstration video, we introduce the Language Grid Toolbox, an open source multilingual communication tool, and two community sites based on the Language Grid Toolbox. The G30 Community Site aims to create a multilingual and multicultural community to accommodate the needs of Japanese and international students. The Pangaea Community Site is used by facilitators of NPO Pangaea located in different countries around the world to communicate in their native languages through machine translation supported multilingual BBS. Copyright 2011 ACM.

© 2011, Springer-Verlag Berlin Heidelberg. The concept of collective intelligence is contributing significantly to knowledge creation on the Web. While current knowledge creation activities tend to be founded on the approach of assembling content such as texts, images and videos, we propose here the service-oriented approach. We use the term service grid to refer to a framework of collective intelligence based on Web services. This chapter provides an institutional design mainly for non-profit service grids that are open to the public. In particular, we deepen the discussion of 1) intellectual property rights, 2) application systems, and 3) federated operations from the perspective of the following stakeholders: service providers, service users and service grid operators respectively. The Language Grid has been operating, based on the proposed institutional framework, since December 2007.

The Language Grid is an infrastructure for enabling users to share language services developed by language specialists and end user communities. Users can also create new services to support their intercultural/multilingual activities by composing language services from a range of providers. Since the Language Grid takes the service-oriented collective intelligence approach, the platform requires the services management to satisfy stakeholders’ needs: access control for service providers, dynamic service composition for service users, and service grid composition and system configurability for service grid operators. To realize the Language Grid, this chapter describes the design concept and the system architecture of the platform based on the service grid.

Since various communities, which use multiple languages, now want to interact in daily life, tools that can effectively support multilingual communication are necessary. However, we often observe that the success of a multilingual tool in one situation does not guarantee its success in another. To develop a multilingual environment that can handle various situations in various communities, existing language resources (dictionaries, parallel texts, part-of-speech taggers, machine translators, etc.) should be easily shared and customized. Therefore, we designed our proposal, the Language Grid, as service-oriented collective intelligence
it allows users to freely create language services from existing language resources and combine those language services to develop new services to meet their own requirements. This chapter explains the design concept and service architecture of the Language Grid, and the approach of user involvement in the collective intelligence activities.

Multiagent-based simulation (MABS) is a very active interdisciplinary area bridging multiagent research and social science. The key technology to conduct truly useful MABS is agent modeling for reproducing realistic behaviors. In order to make agent models realistic, it seems natural to learn from human behavior in the real world. The challenge presented in this paper is to obtain an individual behavior model by using participatory modeling in the traffic domain. We show a methodology that can elicit prior knowledge for explaining human driving behavior in specific environments, and then construct a driving behavior model based on the set of prior knowledge. In the real world, human drivers often perform unintentional actions, and occasionally, they have no logical reason for their actions. In these cases, we cannot rely on prior knowledge to explain them. We are forced to construct a behavior model with an insufficient amount of knowledge to reproduce the driving behavior. To construct such individual driving behavior model, we take the approach of using knowledge from others to complement the lack of knowledge from the target. To clarify that the behavior model including prior knowledge from others offers individuality in driving behavior, we experimentally confirm that the driving behaviors reproduced by the hybrid model correlate reasonably well with human behavior.

With the development of the Internet environments, more and more machine translation tools have become available on the Web, which provides the opportunity for some multilingual communities to use machine translation for communication. However, there exist several major problems in machine translation tools available on the Internet. 1) Translation for specific terms or sentences within communities is always of low quality. 2) Most of the available machine translation tools only provide basic functions of translating sentences, documents or Web pages, and lack necessary communication functions for community users. 3) It is difficult to conduct collaborative work among community users. 4) There is no support for creating customized multilingual environments based on unique community requirements. To address the above issues, we develop the Language Grid Toolbox which enables easy creation of multilingual community sites and customized multilingual environments for communities. For example, Toolbox has the basic function of creating language resources like dictionaries, which are used to combine with atomic translation services to improve the translation quality. Further, since Toolbox is developed as open source software and provides many basic APIs that can be used for communication, customized functions for communities can be easily developed. Several customized communication tools that are extended from Toolbox basic modules have already been implemented by universities and local government. ©2010 IEEE.

Toru Ishida, Kyoto University, suggests that multilingual collaboration approaches need to be adopted to solve the problem of working in multiple cultures. Machine translation has emerged as one of the significant solutions to solve the problem of understanding different languages of the world. Machine translation systems can be useful when they are customized to suit specific needs of different communities. Users need to combine domain-specific texts with machine translators to customize machine translations. They also need morphological analyzers to analyze input sentences that are to be translated. Training machine translators with parallel texts requires dependency parsers, as users will want to use speech recognition/synthesis and gesture recognition in the future. Worldwide collaboration will be needed to generate all the necessary language services for supporting local schools, including students from different countries.

A composite Web service provided as a "cloud" service should make its constituent Web services transparent to users. However, existing frameworks for composite Web services cannot realize such transparency because they lack capability of adapting changes of behaviors of constituents Web services and business rules of service providers. Service Supervision, proposed in the previous work, allows us to flexibly adapt a composite Web service by combining control execution functions which control behavior of running instances of composite Web services. However, much flexibility of the execution control functions sometimes makes it difficult to design adaptation processes due to absence of accumulated know-how such as guidelines. Moreover, it often costs a lot to port adaptation processes to the model of composite Web service to be adapted. To solve the problems, we first organized various adaptation processes based on some previous works. Then we proposed Service Supervision patterns, which consist of typical requirements for adaptation and WS-BPEL processes satisfying the requirements by using execution control functions. The patterns are easy for designers of composite Web services to understand and make it possible to reduce cost to port them to the model of a composite service.

近年急速に活発化しているマルチエージェントシミュレーション研究において,エージェントモデリングは核となる技術である.本論文では,応用領域として交通問題に着目し,参加型モデリングに基づいた個別の運転行動モデルの獲得のために,人間の運転行動を説明するための先験的知識をもとにモデルを構築するための手法を提案する.運転者横断的な先験的知識の利用を行う本手法によって,個別の運転スタイルが再現できることを,実験によって明らかにする.

異言語間のコミュニケーションに,機械翻訳を用いることが試みられている.しかし,機械翻訳を用いた多言語コミュニケーションで共通基盤が問題なく形成されるか否かの評価・検証は行われていない.そこで,本論文では,日本語,韓国語,中国語を第一言語(first language)とする話者を対象に,チャットシステムを介した遠隔コミュニケーションで,共通の言語(英語)を用いる通常の場合と機械翻訳を用いる場合の比較実験を行った.この実験は,3者にそれぞれにランダムに配列された8枚の抽象図形を与え,チャットを通じて同じ図形を同定していくものである.この実験から得られた対話ログを,参照語に着目して分析したところ,以下のことが明らかになった.(1)実験を繰り返すと,通常は,同定の遅かった照合者に合わせた参照表現を選ぶようになるのだが,機械翻訳を用いた場合にはそうはならなかった.(2)実験を繰り返すと,通常は,前回用いられた参照語を使う傾向があるが,機械翻訳を用いるとそうなるとは限らない.(3)実験を繰り返しても,参照語を用いて最短のプロセスで図形を同定できる割合が,共通言語を用いる場合ほど増加しない.これらの現象から,機械翻訳を用いた場合には,自分と他者,あるいは他者と他者のコミュニケーションを正しく理解できないため,共通基盤の成立が困難となっていることが明らかとなった.

Real-time interactive virtual classroom with teleeducation experience is an important approach in distance learning. However, most current systems fail to meet new challenges in extensibility and scalability, which mainly lie with three issues. First, an open system architecture is required to better support the integration of increasing human-computer interfaces and personal mobile devices in the classroom. Second, the learning system Should facilitate opening its interfaces, which will help easy deployment that copes with different circumstances and allows other learning systems to talk to each other. Third, problems emerge on binding existing systems of classrooms together in different places or even different countries such as tackling systems intercommunication and distant intercultural learning in different languages. To address these issues, we build a prototype application called Open Smart Classroom built on our software infrastructure based on the multiagent system architecture using Web Service technology in Smart Space. Besides the evaluation of the extensibility and scalability of the system, an experiment connecting two Open Smart Classrooms deployed in different countries is also undertaken, which demonstrates the influence of these new features on the educational effect. Interesting and optimistic results obtained show a significant research prospect for developing future distant learning systems.

To realize a harmonious multi-agent society, in previous work there are always two separated methods to make agent coordination: individual-local balance perspective and individual-society balance perspective. The two separated methods may bring out the conflicts between local and social performances. To achieve the compatibility between the local and social performances, this paper combines the two perspectives together and makes trade-off between them. With our presented model, the individual, local and social concerns can be balanced well in a unified and flexible manner. The experimental results show that there are often situations in which it is better for the local performance if the globally social performance is improved, and vice versa; thus the two coordination perspectives are not conflictive but often bring out the better in each other. Therefore, it concludes that the local and social performances in multi-agent societies can be compatible if we combine the two coordination methods together. (c) 2008 Elsevier Ltd. All rights reserved.

In this paper, we detect "innovative topics", those that are new and hopefully interesting to the user. We try to expand user interests significantly by letting the user browse those topics.
We first generate user-interest ontologies that allow user profiles to be constructed as a hierarchy of classes where a user interest weight is assigned to each class and instance. Next, we measure the similarity between user interests by using interest weights on their user-interest ontologies and generate user group G(U) that has high similarity to user u. The innovative topics for u are then detected by determining a suitable size of G(U) and analyzing the ontologies in G(U). (C) 2009 Elsevier B. V. All rights reserved.

Customization of systems is costly, but it is necessary to better meet the needs of multi-various tasks and requirements in the fields such as medical care, education, and so on. A machine translation system, for example, can provide a certain level of multilingual communication support if a human translator is absent
the translation system may need to be customized, however, if it is applied to a new application area. To realize easy customization of multilingual communication systems, we have developed the Language Grid Playground, an online language Web service platform which enables easy customization of various multilingual communication tools. We have established a stepwise process of classifying language services into a layered architecture and developing and accumulating lightweight building blocks that constitute communication tool components. Through this stepwise process, we were able to identify useful components in multilingual communication tools which can be flexibly combined to create new Web services. By implementing several tools and actually accumulating building blocks, we identified useful language service components and developed these components as programs which can be deployed as Web services. © 2009 IEEE.

In this paper, we present the concept for collaborative translation, where two non-bilingual people who use different languages collaborate to perform the task of translation using machine translation (MT) services, whose quality is imperfect in many cases. The key idea of this model is that one person, who handles the source language (source language side) and another person, who handles the target language (target language side), play different roles: the target language side modifies the translated sentence to improve its fluency, and the source language side evaluates its adequacy. We demonstrated the effectiveness and the practicality of this model in a tangible way. Copyright 2009 ACM.

Findings on how humans detect cultural differences in cultural pictogram interpretations are reported. An answer web survey was conducted in the United and Japan to collect U.S.-Japan pictogram. Thirty U.S.-Japan pictogram interpretations used as stimuli for human cultural difference study. Three U.S. subjects and three Japanese participated in the study to assess the degree of differences in the thirty pictogram interpretations in the questionnaire. Post-questionnaire interviews conducted to elucidate the reasons behind the human difference detection. The following factors were when humans detect cultural differences in cultural pictogram interpretations: (i) similar or interpretations in the two countries, (ii) or ranking of the interpretations, (iii) or variance of semantics within one country's, (iv) presence of proper nouns (e.g. country ), and (v) positive or negative connotation in the interpretations. Copyright 2009 ACM.

Various types of multilingual collaboration tasks must be performed in the fields of education, medical care, and so on. Members in such fields need support customized for each field. Therefore, multilingual collaboration tools should allow customization to suit the tasks and circumstances. The tools provided by portal sites such as Google and Excite are not flexible enough to solve the problems in various fields because they fail to support customization. Therefore, we have developed the Language Grid Playground: An environment in which it is easy to make customized multilingual tools. The basic idea is to organize language services in a layered architecture and develop light weight building blocks that form collaboration tools by combining services. Our system, which is composed of components designed in this way, makes it easy to create tools customized for various intercultural collaboration fields. As a practical example, we develop a customized tool for the field of education in just 6 man-weeks. It confirms the efficiency of our approach for developing tools.

Pictogram Communication can create an environment where children can develop "bonds" regardless of their differences in language, culture, and physical locations. This paper reveals the communication patterns as extracted from an extensive collection of pictogram messages. More than thirteen hundred messages are assigned to three communication patterns: syntactic description, artistic drawing, and story telling. We examined three patterns of messages by to whom messages were sent. In addition, we observed the development process of such patterns, strategies children seem to apply. Finally, we discuss the influence of cultures on patterns. We found that different mother tongues influence on how children construct messages, while some children reason the cultural differences to create robust messages for intercultural communication.

Companion Modeling is a methodology used to facilitate adaptive management of renewable resources by their users. it is using role-playing games (RPG) and multiagent simulations to validate initial models representing the functioning of complex systems to be managed. In this research, we propose a novel agent model construction methodology in which classification learning is applied to the RPG log data in Companion Modeling. This methodology enables a systematic model construction that handles multi-parameters, independent of the modelers' ability. There are three problems in applying classification learning to the RPG log data: 1) It is difficult to gather enough data for the number of features because the cost of gathering data is high. 2) Noise data can affect the learning results because the amount of data may be insufficient. 3) The learning results should be explained as a human decision making model and should be recognized by the expert as reflecting reality. We realized an agent model construction system using the following two approaches: 1) Using a feature selection method, the feature subset that has the best prediction accuracy is identified. In this process, the important features chosen by the expert are always included. 2) The expert eliminates irrelevant features from the learning results after evaluating the learning model through a visualization of the results. Finally, using the RPG log data from a Companion Modeling case study on rice production in northeastern Thailand, we confirm the capability of this methodology.

Multiagent-based simulations are a key part of several to search research fields. Multiagent-based simulations yield multiagent societies that. well reproduce human societies. acid so are seen as an excellent tool for analyzing the real world. A multiagent-based simulation allows crowd behavior to emerge through interactions among agents where each agent, is affected by the emerging crowd behavior. The interaction between microscopic and macroscopic behaviors has long been considered an important issue, termed the "micro-macro problem", in the field of sociology, but research on the issue is still premature in the engineering domain. are focusing on citywide traffic as a target problem and are attempting to realize mega-scale multiagent-based traffic simulations. While macro-level simulations are popular in the traffic domain. it has been recognized that micro-level analysis is also beneficial. However, there is no software platform that Can realize analyses based on both micro and macro viewpoints due to implementation difficulties. In this paper, we propose a traffic simulation platform that can execute citywide traffic simulations that include driving behavior models. Our simulation platform enables the introduction of individual behavior models while still retaining scalability.

To design socially embedded systems, this paper proposes to learn from human behavior in participatory simulation, where scenario-guided agents and human-controlled avatars coexist in a shared virtual space and jointly perform simulations. To create agent models incrementally, we use machine learning technologies. We characterize agents by using a various combinations of behavior rules instantiated by the user operating his/her avatar. We apply hypothetical reasoning, which offers consistent selection of hypotheses and allows us to start with incompatible behavior rules, for incrementally improving the agent models. Using data obtained during the participatory simulation and the hypotheses including known behavior rules, we can generate an explanation for human behavior.

Avatars are increasingly used to express our emotions in our online communications. Such avatars are used based on the assumption that avatar expressions are interpreted universally among all cultures. This paper investigated cross-cultural evaluations of avatar expressions designed by Japanese and Western designers. The goals of the study were: (1) to investigate cultural differences in avatar expression evaluation and apply findings from psychological studies of human facial expression recognition, (2) to identify expressions and design features that cause cultural differences in avatar facial expression interpretation. The results of our study confirmed that (1) there are cultural differences in interpreting avatars' facial expressions, and the psychological theory that suggests physical proximity affects facial expression recognition accuracy is also applicable to avatar facial expressions, (2) positive expressions have wider cultural variance in interpretation than negative ones, (3) use of gestures and gesture marks may sometimes cause counter-effects in recognizing avatar facial expressions. © Springer-Verlag London Limited 2009.

To realize large-scale socially embedded ambient intelligence systems, this paper proposes a design methodology towards society-centered design. Participatory technologies and multiagent systems are essential in the new system design perspective. Multiagent systems make it possible to test and predict the behavior of socially embedded systems. We have already developed the scenario description language, which describes interaction protocols that link agents to society. We use the virtual space, wherein agents behave under given scenarios, in explaining each step of society-centered design. The process consists of participatory simulation, where agents and human-controlled avatars coexist in virtual space to jointly perform simulations, and augmented experiment, where an experiment is performed in real space by human subjects, scenario-controlled agents, and human extras. For realizing realistic interactions between agents and humans during participatory simulations, an agent model that can reproduce human-like agent behaviors is needed. We show a direction for agent modeling based on learning from humans in actual application environments.

<p>Findings on how humans judge cultural differences in cross-cultural pictogram interpretations are reported. An open-answer web survey was conducted in the United States and Japan to collect English‐Japanese pictogram interpretations for 120pictograms used in an email system.30 pictograms with possible cultural differences were selected as stimuli for investigating human cultural difference judgment criteria. Six human subjects, three Americans and three Japanese, assessed the degree of cultural differences in 30 U.S.‐Japan pictogram interpretations using a questionnaire. Interviews were conducted to elucidate the reasons behind the judgment. The following factors were considered when judging cultural differences in cross-cultural pictogram interpretations: (i)similarity or difference of interpretations between two countries, (ii)percentage or ranking of interpretations, (iii)conformity or variance of interpretations within one country, (iv)presence of contrasting pair of proper nouns such as country names, and(v)positive or negative connotations underlying the two countries' interpretations.</p>

We propose an approach to overcome cultural barriers, specifically pragmatic misunderstandings about the intention of the speaker by annotating domain specific language resources according to an ontology based on speech acts. These annotations are capable to account for indirect speech acts as well as for phenomena of mitigation and reinforcement. The presented architecture preserves the original structure of the resource by providing metadata specified using RDF. We illustrate our purpose by a manually annotated corpus designed to facilitate interaction with foreign patients in Japanese hospitals. © 2008 IEEE.

Configuring a composite Web service by setting endpoints reduces the cost of development, but raises the probability of a request message triggering runtime execution failures. Previous works oil validation of composite Web services are riot useful because the application developer cannot modify atomic/composite services and the specifications needed for validation are riot always available. Therefore, in this paper, we address two issues: predicting the executability of composite Web services for each request message, and acquiring input specifications to improve the prediction. To resolve these issues, we model atomic/composite services in a formal specification. Moreover, we apply constraint acquisition algorithm to acquire input specifications of atomic Web services. We conduct all experiment in which the proposed method is applied to a composite Web service in practical use. The result shows that our method call detect almost all messages that will trigger execution failure at a rather early stage of specification acquisition.

Recently, large-scale distributed multiagent systems consisting of one million of agents have been developed. When agents are distributed among multiple servers, both the computational and interaction cost of servers must be considered when optimizing the performance of the entire system. Multiagent systems reflect the structure of social communities and artificial networks such as the Internet. Since the networks possess characteristics common to the 'small world' phenomenon, networks of agents oil the systems can be considered as small worlds. In that case, communities, which are the sets of agents that, frequently interact with each other, exist in the network. Most previous works evaluate agents one by one to the most appropriate agent, to be moved to a different server. If the networks of agents are highly clustered, previous works divide the communities when moving agents. Since agents in the same community often interact with each other, this division of communities increases the interaction cost among servers. We propose community-based load balancing (CLB), which evaluates the communities to select the most appropriate set of agents to be moved. We conducted simulations to evaluate out. proposed method according to the network of agents. Our simulations show that when the clustering coefficient is close to 1.0 the interaction cost with CLB can be approximately 30% lower than that with previous works.

A social law is a restriction on the set of strategies available to agents [1]. Each agent can select some social strategies in the operation of the systems, however, the social strategies of different agents may collide with each other. Therefore, we need to endow the global social laws for the whole system. In this paper, the social strategy is defined as the living habits of agent, and the social law is the set of living habits which can be accepted by all agents. This paper initiates a study of evolving social strategies of individual agents to global social law of the whole system, which is based on the hierarchical immediate diffusion interaction from superior agents to junior ones. In the diffusion interactions. the agents with superior social position can influence the social strategies of junior agents, so as to reduce the social potential energy of the system. The set of social strategies with the minimum social potential energy can be regarded as the global social law.

Many research projects have studied various aspects of smart environments including smart rooms, home, and offices. Few projects, however, have studied smart urban public spaces such as smart railway stations and airports due to the lack of an experimental environment. We propose virtual cities as a testbed for examining the design of smart urban public spaces. We developed an intelligent emergency guidance system for subway stations and used the virtual subway station platform to analyze the effects of the system. This experience allows us to argue that simulations in virtual cities are useful to pre-test the design of smart urban public spaces and estimate the possible outcome of real-life scenarios. © 2009, IGI Global.

There are always two kinds of forms in agent social law diffusion, which are local interaction and non-local coordination. This paper provides an integrated model to make a tradeoff between local diffusion and non-local diffusion. In the local interaction, each agent often interacts with a small set of social 'neighbors' and imitates the locally most authoritative law; in the non-local coordination, agents may coordinate with others that are not in the neighboring places for some tasks, and will negotiate about their social laws according to the coordination strategy. With the presented model, the diffusions between local interaction and non-local coordination can be harmonized, and the diffusion impacts of different strategies in non-local coordination can also be addressed well. (c) 2006 Elsevier Ltd. All rights reserved.

Flexibility, adaptation and distribution have been regarded as major challenges of modern interorganizational workflow. To address these issues, this paper proposes an interorganizational workflow execution framework based on process agents and ECA rules. In our framework, an interorganizational workflow is modeled as a multiagent system with a process agent for each organization. The whole execution is divided into two parts: the intra-execution, which means execution within a same organization, and the inter-execution, which represents interaction between organizations. For intra-execution, we use the method of transforming the graph-based local workflow into block-based workflow to design general ECA rules. ECA rules are used to control internal state transitions and process agents are used to control external state transitions of tasks in the local workflows. Inter-execution is realized by process agent interaction protocols. The proposed approach can provide flexible execution of interorganizational workflow with distributed organizational autonomy and adaptation. A case study of offshore software development is illustrated for the proposed approach.

翻訳機を介した円滑な相互理解を実現するためには，翻訳精度の向上とともに，翻訳機を介したコミュニケーションの性質に合わせた支援が重要である．我々は相互理解の構築に重要な参照（reference）に注目し，2 人の異言語ユーザが共通言語（英語）を用いて参照内容を同定するプロセスと翻訳機を用いて母国語で同定するプロセスを比較する実験を実施した．実験では異言語ユーザがペアとなりチャットシステムを介して同じ抽象図形のマッチングを繰り返した．実験より得た対話ログを分析し先行研究から導出した仮説を検証したところ，翻訳機を介した対話に以下の特徴があることが分かった．1．参照語による同定が難しい；一方が呈示した参照語を他方が受理するという最短のプロセスで同定できるケースが少なかった．2．互いに質問をしたり確認することは少ない；抽象図形の様々な特徴を次々に呈示することで伝えようとしていた．3．意味を保持したフレーズの省略（simplification）が少ない；相互理解が構築された表現を再び伝える場合は，短縮せずにそのまま伝えるか，もしくは特徴的な情報のみ伝えていた（narrowing）Reaching mutual understanding over lean media using low quality machine translation is a difficult task. To support such communication, it is important to incorporate functions that reflect actual communication pattern as well as improving its translation quality. In this study, we focused on referring behavior, which is closely related to mutual understanding. We compared referring behavior in regular communication with referring behavior in machine translation-mediated communication. The analysis results indicate that in machine translation-mediated communication, 1. it is difficult to shorten a referring phrase to a noun phrase; when one issued a noun phrase, the other had trouble in accepting the noun phrase. 2. one continuously presented information from various angles until the other got the idea of what was meant. 3. one tend to repeat the same phrase or narrow the expression to distinctive part of the referent.

There were always two separated methods to make agent coordination: individual-local balance perspective and individual-society balance perspective. The first method only considered the balance between individual agents and their local neighbors
the second method only considered the balance between individual agents and the whole multiagent society. However, in reality, the agents will be diffused by their local neighbors as well as influenced by their social contexts simultaneously
therefore, it is necessary to deal with the social performance as well as local performance. To address such problem this paper presents an agent coordination method in an integrative model where we combine the two perspectives together and make trade-off between them. With our presented model, the individual, local and social concerns can be balanced well in a unified and flexible manner. Moreover, the experimental results show that there are often situations in which the two coordination perspectives aren't conflictive but often bring out the better in each other. © 2007 IFAAMAS.

This paper studies the patterns of cultural differences observed in pictogram interpretation. We conducted a 14-month online survey in the U.S. and Japan to ask the meaning of 120 pictograms used in a pictogram communication system. A total of 935 respondents in the U.S. and 543 respondents in Japan participated in the survey to submit pictogram interpretations which added up to compose an average of 147 English interpretations and 97 Japanese interpretations per pictogram. Three human judges independently analyzed the English-Japanese pictogram interpretation words, and as a result, 19 pictograms were found to have culturally different interpretations by two or more judges. The following patterns of cultural differences in pictogram interpretation were observed: (1) two cultures share the same underlying concept, but have different perspectives on the concept, (2) two cultures only partially share the same underlying concept, and (3) two cultures do not share any common underlying concept. © Springer-Verlag Berlin Heidelberg 2007.

Our proposed "Language Grid" infrastructure supports multilingual communication by combining in new way language resources, such as machine translators, morphological analyzers, and dictionaries specific to user communities. We developed the Language Grid as a language infrastructure on the Internet. The Language Grid enables user communities to combine two or more machine translators and their community dictionaries by workflows, and to easily create new multilingual services specific to the communities. Because the quality of language services is not often defined, however, we need to confirm that the created multilingual service is really useful. We need to extend the process of general usability testing to the multilingual environment. For example, cooperation between user communities and language grid providers can significantly improve the accuracy of machine translation: it turns out that machine translations can be useful for interactive communication in the field of inter-cultural collaboration. © Springer-Verlag Berlin Heidelberg 2007.

To retrieve pictograms having semantically ambiguous interpretations, we propose a semantic relevance measure which uses pictogram interpretation words collected from a web survey. The proposed measure uses ratio and similarity information contained in a set of pictogram interpretation words to (1) retrieve pictograms having implicit meaning but not explicit interpretation word and (2) rank pictograms sharing common interpretation word(s) according to query relevancy which reflects the interpretation ratio. © Springer-Verlag Berlin Heidelberg 2007.

機械翻訳を介したコミュニケーションを通じて相互理解を実現するためには,翻訳精度の向上とともに相互作用性の向上が重要である.我々は機械翻訳に対するユーザの適応行動の1つである原文の書き換えに注目した.本論文では,ユーザが母国語だけを用いて原文の書き換え作業を行う方法として折り返し翻訳を検討し,折り返し翻訳を用いてユーザが書き換え作業をする際の作業量を減らす支援方法を考案する.本研究でユーザの折り返し翻訳作業に関する実験を実施,分析した結果,以下の知見を得た.1)母国語に関する知識が豊富なユーザほど機械翻訳に容易に適応でき書き換え作業量が少なかった.2)ユーザに事前に「良い翻訳結果を得るためのルール集合」の教示を行うと,母国語に関する知識が豊富でないユーザも機械翻訳に容易に適応できるようになり,書き換え作業量が大幅に減った.3)ただし,原文をどのように変更すべきかを明示しない「操作自由型ルール」に対する教示効果は薄く,これらのルール獲得にかかる書き換え作業量は大きく減少しなかった.原文をどのように変更すべきかを明示した「操作指示型ルール」に対する教示効果は高く,これらのルール獲得にかかる書き換え作業量は大きく減少した.4)ルールの教示は,母国語に関する知識が中位のユーザに最も効果的であった.

東アジアを中心に，機械翻訳を用いて議論をするコミュニティが増加している．しかし，機械翻訳を用いた対話には誤訳ノイズが混入するため，ユーザ間の相互理解を構築することは難しい．我々は，機械翻訳を介した対話で大量に発生する思い違いの問題に焦点を当て，機械翻訳を介した対話の特徴と思い違いの関係について分析，考察した．その結果，思い違いが大量に発生する異国ユーザ間の返信メッセージは，内容をとらえずメッセージの部分的なところでつながる傾向が強いことが分かった．そこで，これらの対話の特徴をふまえ，誤訳による思い違いを機械的に検知する手法を提案した．提案手法では，通常のスレッド（syntactic thread）と語彙的結束性に基づいたスレッド（semantic thread）の差異を計測し，この差異が大きければユーザ間の話題に関する思い違いが生じやすいと判断するものである．実際に対話データを用いて提案手法を検証したところ，提案手法によって計測した思い違いと実際の思い違いの間には有意な正相関が観測され，手法の有効性が示された．Multilingual communities using machine translation to overcome language barriers are showing up more and more frequently. However, when a large number of translation errors get mixed into conversation, it becomes difficult for users to fully understand each other. In this paper, we focus on misconceptions found in high volume in actual online conversations using machine translation. We first examine how misconceptions occurred by delving into 1106 messages exchanged on BBS using machine translation. The analysis results indicate that when a user responds via machine translation, he/she tends to respond to short phrases of the original message and tends to trip on the wording of the original message. Next, based on the analysis results, we propose a method that automatically measures the likeliness of each dialogue including misconceptions. The proposed method assesses the likeliness of each dialogue including misconceptions by calculating the gaps between the regular discussion thread (syntactic thread) and the discussion thread based on lexical cohesion (semantic thread). Verification results show significant positive correlation between actual misconception frequency and the syntax-semantic gap, which indicates the validity of the method.

In this paper, we describe new methodologies for reinforcing the social consensus building of the multicultural coexistence assistance using participatory simulation in civil societies. In this methodology, both human and intelligent software agents participate in a participatory gaming simulation. This simulation, which has mutual interaction between humans and agents, can achieve the gathering and analyzing of beneficial information to solve problems in multicultural civil societies. This type of simulation will help to provide knowledge for the formulation of social consensus, and to create new opportunities for discussion with people with different cultures and social backgrounds. © 2006 IEEE.

To realize large scale socially embedded systems, this paper proposes a multiagent-based participatory design that consists of steps called 1) participatory simulation, where scenario-guided agents and human-controlled avatars coexist in virtual space and jointly perform simulations, and 2) augmented experiment, where an experiment is performed in real space by human subjects and scenario-guided extras. In this methodology, we use production rules to describe agent models for approximating users, and multiagent scenarios to describe interaction models among services and their users. To learn agent and interaction models incrementally from simulations and experiments, we establish the participatory design loop with deductive machine learning technologies. Copyright 2006 ACM.

Inter-organizational projects involve the creation, modification and management of content. Unless carefully handled, these overheads can came the loss of mutual understanding. In this paper, we propose a content management approach for inter-organizational projects that uses the e-mail metaphor
the functions of creating, modifying, and managing content are represented as e-mail messages that are sent to project members automatically. The e-mail metaphor allows us to communicate the creation, modification and management of content explicitly across organizations. It promotes mutual understanding in interorganizational projects. We have developed a content management system that uses the e-mail metaphor. When one project member adds content to the project, the system informs this event to the other project members. When the project manager manages the content by grouping, commenting or semantic annotation, the system informs to project members of the operation by e-mail metaphor such as creating a mailbox for a new group. © 2006 IEEE.

Though research on agents that interact with humans via voice or text has been intensively conducted, agents that can interact with more than two persons in parallel have not been studied well. To enable an agent to interact with multiple people, we propose a method to assign multiple scenarios to one agent, in which each scenario describes an interaction protocol between the agent and one person. Obviously, coordination among multiple scenarios is required to avoid conflicts in actions. For example, one agent cannot execute both walking and sitting actions simultaneously. However, what is more important is that a coordination policy, a way of specifying how to manage conflicts among multiple actions, must be introduced. In this paper, we introduce a coordination scenario that avoids conflicts in actions and coordinates scenarios according to a coordination policy. The coordination scenario, which controls the execution of interaction scenarios by communication, is generated by a coordination policy for solving conflicts. When the coordination scenario receives a request to execute actions from an interaction scenario, it checks whether the actions will trigger conflicts and sends an order not to execute them if conflicts will occur. The coordination scenario interworks concurrent interaction scenarios by executing this process repeatedly. © Springer-Vorlag Berlin Heidelberg 2006.

In massive agent system, there are many diffusion processes among agent social strategies which take place concurrently. A social law is a restriction on the set of strategies available to agents [1]. All agents will trend to select an identical social strategy in the agent social law evolution, which can be called as the phenomenon of unification trend. This paper presents a model for the concurrent social strategy diffusion with unification trend. With the model, an agent's social strategy is influenced not only by the diffusion that bear on itself, but also by concurrent diffusion processes that bear on other agents; and, an agent will incline to the average social strategy of the whole system which can make the system be more unified.

調達取引に代表される商取引では，仕様書・仕様データに付随する情報を取引相手と如何に共有できるかが成否の鍵となる．我々は電子調達における取引過程のコンテクストをシミュレーション実験により得て分析し，仕様データに付随する情報が調達活動のコンテクストの中で利用され，知識と経験が共有される過程を考察した．この考察結果に基づいて，本稿では企業の成長過程における協業モデルを整理した上で，企業間協業モデルと仕様情報の相互理解や知識経験の共有とが，どのように対応づけられるかを考察した．

This paper proposes a new e-procurement model for a large number of buyers and sellers interacting via the Internet. The goal of e-procurement is to create a satisfactory match between buyers' demand and sellers' supply. From our real-world experience, we view e-procurement as a process of negotiation to increase the matching quality of two corresponding specifications: one for buyers' demand and another for sellers' supply. To model scalable e-procurement, we propose a co-adaptive matchmaking mechanism using mutual relevance feedback. In order to understand the nature of the mechanism, we have developed two types of software agents, called e-buyers and e-sellers, to simulate human buyers and sellers. Multiagent simulation results show that the matching quality is incrementally improved if agents adaptively change their specifications. A realistic example is also provided to discuss how to extend our simulation to real-world e-procurement infrastructure. © Springer-Verlag Berlin Heidelberg 2005.

Multilingual communities using machine translation to overcome language barriers are showing up with increasing frequency. However, when a large number of translation errors get mixed into conversation, it becomes difficult for users to fully understand each other. In this paper, we focus on misconceptions found in high volume in actual online conversations using machine translation. By comparing responses via machine translation and responses without machine translation, we extract two response patterns, which may be strongly related to the occurrence of misconceptions in machine translation-mediated communication. The two response patterns are that users tend to respond to short phrases of the original message and tend to trip on the wording of the original message when responding via machine translation.

We consider a framework where agents perform trust formation on behalf of human users in open environment like pervasive computing or the Internet. In this paper, an automated trust negotiation framework [1, 2, 3, 4] is proposed to support such trust agents, and a protocol is presented. In the framework, the preference for the disclosure of credentials and policies are explicitly included in contrast to ordinary ATN frameworks studied in network authentication domain. The proposed protocol is based on dynamic programming and successfully performs the exchange of credentials and policies without unnecessary disclosure of them. © 2005 IEEE.

In this paper, we present the evacuation simulation system that is the combination of a virtual city and a crisis management simulation. The system allows users to become virtual evacuees in an evacuation simulation to learn about crowd behavior. In the experimental use of the system, we found that the synergic effects between a bird's-eye and a first-person views in learning emergency escaping behaviors. Based on this result, we designed a novel communication system that allows a remote leader to guide escaping crowds in an emergency situation. We deployed our prototype in the Kyoto Station.

Digital Cities have been developed all over the world to provide regional city information. This paper proposes a new approach to integrate existing digital city systems via a universal mobile interface. The universal mobile interface integrates various media representations in connected digital city systems representing the same real city with audio semantic annotations. It is also applicable for a mobile use without developing a specific mobile interface. The demo system that connects three heterogeneous digital cities, namely, a 3D modeled virtual space, image-based virtual space, and map-based information space, successfully showed the connectivity of the universal interface. The current system provides the following functions: 1) a single and simple interface to navigate users into multiple digital cities, 2) a simultaneous walk-though into the connected digital cities, 3) audio descriptions of scenes using semantic annotation of objects in the connected digital cities, 4) prioritizing semantic annotation according to users' needs.

Future computing systems interact with a large number of users moving around buildings and streets. In this paper, we propose an example of such systems and how to evaluate ubicomp systems equipped with a large-scale physical environment that includes a large number of people inside. In our emergency guidance system, off-site guiding staff monitors a crowded large-scale public space to understand its situation, and instruct on-site guiding staff how to guide crowds effectively. Our system tracks and synthesizes the public space to enable the off-site staff to grasp it, and support communication between the on-site and the off-site staffs. Because it is not affordable to use the physical public space and a lot of human subjects to evaluate the system, we used our social interaction platform to simulate our guidance system. We could successfully construct simulations, in which the crowds are replaced with social agents in the virtual public space.

Domain-specific Web search engines are effective tools for reducing the difficulty experienced when acquiring information from the Web. Existing methods for building domain-specific Web search engines require human expertise or specific facilities. However, we can build a domain-specific search engine simply by adding domain-specific keywords, called "keyword spices," to the user's input query and forwarding it to a general-purpose Web search engine. Keyword spices can be effectively discovered from Web documents using machine learning technologies. This paper will describe domain-specific Web search engines that use keyword spices for locating recipes, restaurants, and used cars.

Real-time search provides an attractive framework for intelligent autonomous agents, as it allows us to model an agent's ability to improve its performance through experience. However, the behavior of real-time search agents is far from rational during the learning (convergence) process, in that they fail to balance the efforts to achieve a short-term goal (i.e., to safely arrive at a goal state in the present problem solving trial) and a long-term goal (to find better solutions through repeated trials). As a remedy, we introduce two techniques for controlling the amount of exploration, both overall and per trial. The weighted real-time search reduces the overall amount of exploration and accelerates convergence. It sacrifices admissibility but provides a nontrivial bound on the converged solution cost. The real-time search with upper bounds insures solution quality in each trial when the state space is undirected. These techniques result in a convergence process more stable compared with that of the Learning Real-Time A* algorithm. (C) 2003 Elsevier Science B.V. All rights reserved.

In this paper, an information navigation agent is proposed as a new interface for obtaining Web information in a mobile environment. A central task of information retrieval systems developed up to now has been to narrow down a great deal of information by displaying search results to a user, followed by refinement of queries on the basis of these results. These have been methods suitable for use in a desktop environment. In contrast, the authors' agent can navigate toward incrementally better information by generating approximating solutions based on constraints on the attributes of the search object and obtaining further information by dialogue with the user. A scheme in which the agent resolves search conditions by using association rules between keywords and adds new search conditions has been proposed. In addition, since a navigation failure occurs if the association rules are used inappropriately, refining the association rules by using statistical tests and graphic analyses has been proposed and its effectiveness has been verified.

本稿の目的は仮想としにおける周囲の人や物体など環境によって引き起こされる対話を扱う対話エージェントの提案である.従来の対話エージェントでは主に前後の文脈に存在するタスク指向対話を扱っているが,本稿では,仮想都市における重要な対話の一つである,環境の属性や非言語的合図を手がかりとする環境指向型対話に着目した.環境指向型対話はユーザの注視・対象の特徴からなる手がかりを条件とし,エージェントの行動を出力するルールを用いてモデル化される.また,我々は仮想都市において環境指向型対話を扱うエージェントをシナリオ記述言語によって実装した。

Current research issues on web services have come to center around flexible composition of existing services. Under the initiative of industry, flexible composition framework has been developed on a workflow model where flow of the processes and bindings among services should be known beforehand. In short, its framework realizes flexible composition within the available services that are not widely opened. This paper focuses on two limitations of current web service composition. One limitation is that it's hard to represent multi-agent scenarios consisting of several concurrent processes because it is based on a workflow model. The other limitation is that once composed, web service cannot be reused or transferrd for other requesters, because there is no function to put the semantics on composite web service. To overcome these limitations, we have developed scenario description language Q, which enables us to realize a web service composition reflecting a multi-agent's context not as a workflow but a scenario. Furthermore, we developed a system that translates multi-agent scenario to DAML-S, and that registers the translated DAML-S as a new Web service. We also discuss the availability of our system for designing an application to C-Commerce and Digital Cities.

社会的エージェントが人間関係に与える影響を調べる第1歩として，エージェントがバランス理論に従って人間関係に影響を与えうるかを調べる実験を行った．エージェントが人間の意見に対して賛成や反対を表明することで，人間がエージェントに対して好意や敵意を抱くかを確かめた．次に，人間エージェント間の関係と人間関係の間にバランス理論が働くかを実験した．実験結果から，人間同士が会話をしない状況ではエージェントが人間関係に影響を与えることが分かった．しかし，人間同士で少しでも会話をすると，エージェントは人間関係に影響を与えられなくなった．会話分析の結果から，人間同士の会話による同調がエージェントの賛成や反対の効果を失わせることが分かった．We conducted an experiment to determine whether agents could influence human relations along the lines of the balance theory.We confirmed that our agent could win favorable feelings from subjects by showing an agreeing attitude to them and,conversely, unfavorable feelings by showing a disagreeing attitude.Next, we validated the balance theory as a rule governing both agent-human relations and human relations.In our experiment,we found that the agent could influence subjects' relations providing that they did not have a conversation.As a result of conversation analysis,we conclude that the subjects' tendency to sympathize with others through conversation eliminated the influence of the agent's agreeing or disagreeing attitude.

The Q language-which describes interaction scenarios between agents and users based on agent external roles-provides an interface between computing professionals and scenario writers.

本研究では,Web特有のハイパリンク解析からWebコミュニティを抽出するKleinbergのHITS(Hyperlink-Induced Topic Search)アルゴリズムの問題点を分析し,改良を行った.分析のために,HITSアルゴリズム動作過程可視化ツール(LinkViewer)を開発した.分析結果から得られたHITSアルゴリズムの問題点は,base集合には本来のトピックには全く関係のないページが密なリンク構造をもつ場合,本来のトピックに見合ったWebコミュニティ(authorityとhub)を抽出することができないことであった(topic drift問題).これに対し,我々は,リンク解析のみに焦点を絞り以下の修正法を提案した.1)固有値計算を行う際にroot部分空間に射影を行う手法,2)root集合に含まれる複数のページとリンク関係をもつページのみをbase集合から取り出して反復計算する手法,及び2)に1)を組み合わせて適用した手法である.結果として,比較的少ない計算量で,いずれのトピックでもtopic driftの問題が是正され,リンク情報を用いたHITSアルゴリズムの改善に成功した.

Traditional crowd simulators are based on a simple numerical analysis of inputs such as the positions of people and structures
they do not consider leadership. Since leaders (in terms of evacuations) are present in many real-world situations, the validity of evacuation simulations can be increased through their introduction. We assess the results of a real-world evacuation experiment to develop more realistic scenarios for simulation. The simulations produced by the improved scenarios are found to closely mimic the experimental results. This work shows that scenario-based agent systems such as FlatWalk and FreeWalk offer excellent promise in simulating evacuations more realistically.

As a platform for community networks, public information spaces that mirror the city metaphor are being developed around the world. The aim of digital cities is to pursue a future information space for everyday urban life, unlike the creation of new businesses which is the current obsession of the Internet. We started the basic research project called "Universal Design of Digital City," a five year project established in 2000, a part of the Core Research for Evolutional Science and Technology (CREST) run by the Japan Science and Technology Corporation (JST). The objective of this project is to construct digital cities as the infrastructure that encourages the participation of all people, including the disabled and the aged. We will develop basic technologies for the universal design, focusing on &apos;sending information,&apos; &apos;receiving information,&apos; and &apos;participation.&apos; This paper introduces some of various experiments such as crisis management, environmental learning, and shopping street navigation. Digital cities are not imaginary since they correspond to the physical urban spaces in which we live. Basic technologies including perceptual information infrastructure and social agents are being developed for connecting digital and physical cities.

As opportunities for international collaboration and crosscultural communication among people from heterogeneous cultures increase, the importance of electronic communication support is increasing. To support cross-cultural communication, we believe it is necessary to offer environments in which participants enjoy conversations, which allow them to share one another's background and profile visually.
We believe that the following three functions are important: (1) showing topics based on participants' profiles and cultural background; (2) lifesized, large-screen interface; and, (3) displaying objects which show feelings of identify. In this paper, we discuss the implementation and the empirical evaluation of two systems that were designed to support cross-cultural communication in the real world or between remote locations.
From the empirical evaluation of these systems, we conclude that these systems add new functionality to support conversation contents, which may be especially useful in a cross-cultural context where language skills are an issue, and this type of environment may be especially useful in a pre-collaboration context.

本論文では, モバイル環境におけるWeb情報の活用に適した新しいインタフェースとして, 情報ナビゲーションエージェントを提案する. これまでの情報検索システムは, 検索結果のユーザへの表示とそれに基づくクエリの変更を通して多くの情報をいかに絞り込んでいくかが中心的課題であり, デスクトップ環境での利用に適した方法であった. それに対して我々のエージェントは, 検索対象の属性の制約条件を用いて近似解を生成し, ユーザとの対話を通して情報を得ることにより, 漸近的により良い情報へとナビゲートしていくことができる. その際に, エージェントがキーワード間の連想ルールを用いて検索条件の解決と新たな検索条件の追加を行う手法を提案した. また, 連想ルールを不用意に用いるとナビゲーションの失敗を導くため, 統計的検定とグラフ構造の解析を用いた連想ルールの精錬を行うことを提案し, その有効性を確認した.

In robot navigation, one of the important and fundamental issues is to find positions of landmarks or vision sensors located around the robot. This paper proposes a method for reconstructing qualitative positions of multiple vision sensors from qualitative information observed by the vision sensors, i.e., motion directions of moving objects. In order to directly acquire the qualitative positions of points, the method proposed in this paper iterates the following steps: 1) observing motion directions (left or right) of moving objects with the vision sensors, 2) classifying the vision sensors into spatially classified pairs based on the motion directions, 3) acquiring three point constraints, and 4) propagating the constraints. Compared with the previous methods, which reconstruct the environment structure from quantitative measurements and acquire qualitative representations by abstracting it, this paper focuses on how to acquire qualitative positions of landmarks from low-level, simple, and reliable information (that is, "qualitative"). The method has been evaluated with simulations and also verified with observation errors.

A digital city is a social information infrastructure for urban life (including shopping, business, transportation, education, welfare and so on). We started a project to develop a digital city for Kyoto based on the newest technologies including cooperative information agents. This paper presents an architecture for digital cities and shows the roles of agent interfaces in it. We propose two types of cooperative information agents as follows: (a) the front-end agents determine and refine users' uncertain goals, (b) the back-end agents extract and organize relevant information from the Internet, (c) Both types of agents opportunistically cooperate through a blackboard. We also show the research guidelines towards social agents in digital cities; the agent will foster social interaction among people who are living in/visiting the city.

This paper describes a framework to construct interface agents with example dialogs based on the tasks by the machine learning technology. The Wizard of Oz method is used to collect example dialogs, and a finite state machine-based model is used for the dialog model. We implemented a Web-based system which includes these functions, and empirically examined the system which treats with a guide task in Kyoto through the experimental use.

This paper proposes a Distributed Vision System (DVS), which is an intelligent infrastructure for mobile robots consisting of many vision agents (VAs) embedded in an environment. The system monitors the environment from various viewing points with the VAs, maintains the dynamic environment models, and provides various information to robots. Based on this concept, we have developed a prototype of the DVS which consists of sixteen vision agents and simultaneously navigates two robots in a model town.

研究奨励賞（14件中1位）

As a platform for community networks, information spaces using the city metaphor are being developed in worldwide. This paper compares the trials of digital cities. Four digital cities, those of America Online, Amsterdam, Helsinki and Kyoto, are introduced. It is interesting to note that each digital city has a different goal: to explore a vertical market, a public communication space, a next generation metropolitan network, and a social information infrastructure for the 21st century, respectively. Their different services, system architectures, and organizations result from the different goals. Digital cities will change together with the advance of computer and network technologies. No digital city can remain at its current status. This paper reviews those digital cities to have a better understanding of their current status and future.

In this paper, we present an organizational overview of the NTT Open Laboratory's Digital City Project.

This paper proposes a method for reconstructing qualitative positions of landmarks by using geometrical constraints. The acquisition process iterates the following steps: (1) observing relations among landmarks and motion directions of moving objects, (2) classifying the landmarks into a pair of landmark sets which are classified with a straight line, (3) acquiring three-point constraints from the pair, and (4) propagating the constraints. We have applied this method to acquisition of qualitative positions of multiple vision sensors in a distributed vision system, and evaluated it with simulations. © 2000 Scripta Technica.

The Wizard of Oz (WOZ) method has been used in natural language processing area especially for evaluating and getting corpus for dialogue design. Furthermore in reference, this method was used to formalize the assumptions about what could be implemented in the interface agent system. The unique feature of applying WOZ is simulating human-computer interaction as distinct from human-human interaction. Our new approach is applying WOZ method to learning interface agent. The role of wizard is not only simulating unimplemented functions instead of system, but also assisting system to learn to fi...

A meeting environment for casual communication in a networked community, FreeWalk provides a 3D common area where everyone,can meet and talk freely. FreeWalk represents participants as 3D polygon pyramids, on which their live video is mapped. Voice volume remains proportional to the distance between sender and receiver. For evaluation, we compared communications in FreeWalk to a conventional desktop videoconferencing system and a face-to-face meeting.

In order to understand cognitive aspects of autonomous robots, it is fruitful to develop a mechanism by which the robot autonomously analyzes physical sensor data and construct a state space. This paper proposes a coherent approach to constructing such a robot oriented state space by statistically analyzing sensor patterns and rewards given as the result of task executions. In the state space construction, the robot creates sensor pattern classifiers called Empirically Obtained Perceivers (EOPs) which, when combined, represent internal states of the robot. A novel feature of this method is that the EOP directs attention to select necessary information, and the state space is obtained with the attention control mechanism using EOPs. We have confirmed that the robot can effectively construct state spaces through its vision sensor and execute a navigation task with the obtained state spaces in a complicated simulated world.

This paper proposes the concept of digital cities as a social information infrastructure for urban life (including shopping, business, transportation, education, welfare and so on). We propose the three layer architecture for digital cities: a) the information layer integrates both WWW archives and real-time sensory information related to the city, b) the interface layer provides 2D and 3D views of the city, and c) the interaction layer assists social interaction among people who are living/visiting in/at the city. We started a three year project to develop a digital city for Kyoto, the old capital and cultural center of Japan, based on the newest technologies including GIS, 3D, animation, agents and mobile computing. This paper introduces the system architecture and the current status of Digital City Kyoto.

In this paper, we develop a formalism called a distributed constraint satisfaction problem (distributed CSP) and algorithms for solving distributed CSPs. A distributed CSP is a constraint satisfaction problem in which variables and constraints are distributed among multiple agents. Various application problems in Distributed Artificial Intelligence can be formalized as distributed CSPs. We present our newly developed technique called asynchronous backtracking that allows agents to act asynchronously and concurrently without any global control, while guaranteeing the completeness of the algorithm. Furthermore, we describe how the asynchronous backtracking algorithm can be modified into a more efficient algorithm called an asynchronous weak-commitment search, which can revise a bad decision without exhaustive search by changing the priority order of agents dynamically. The experimental results on various example problems show that the asynchronous weak-commitment search algorithm is, by far more, efficient than the asynchronous backtracking algorithm and can solve fairly large-scale problems.

We have developed a system, called Silhouettell, that provides awareness support for real-world encounters. Silhouettell uses a large graphics screen. People&#039;s locations (who and where) are projected as shadows on the screen. The feedback from the shadows allows people to naturally know each other. Silhouettell also selects and presents topics common to two to or more people to make conversations easier to start. The current implementation uses World Wide Web (WWW) pages as the material describing the common topics. Experiments with three users are reported to show how Silhouettell works in practice. We also examined where the system would be best used by polling people from various nations.

We have developed a desktop meeting environment named FreeWalk that supports casual meetings in a 3-dimensional (3D) virtual shared space, community common. Tools for human communicadons are always required to run at a practical speed in a widely-used non-luxury hardware environment. However, in the previous researches on 3D conferencing systems, such as VRML-based systems, this aspect has been sometimes neglected. Though we need a research testbed for investigating human-human interactions, researchers tend to concentrate on advanced 3D technologies. In the area of videogames, on the other hand, where 3D virtual spaces are also provided, technologies used there successfully achieve an attractive presentation under severe hardware constraints. Therefore, in the development of FreeWalk. we have applied videogame technologies especially to friendly interface and efficient visualization, so to realize casual communications in a network.

Mobile computing has been mainly utilized for point-to-point communication services, such as E-mail or FAX among people. To advance this field, the challenge is to encourage group communication by providing information that encourages community formation. As the first step towards this goal, we experimentally implemented the community viewer, which dynamically visualizes the communication interaction among people in the community. To design the community viewer, we introduced (1) the party room metaphor, which provides a virtual place for representing various community activities, and (2) the reflector icon, which reflects the activity of the corresponding individual while protecting his/her privacy. The static relationship among people and their dynamic activities are displayed in the spatial arrangement of reflector icons in the party room. We report our experiments on implementing and testing the community viewer in an international conference using 100 personal digital assistants.

This paper investigates real-time bidirectional search (RTBS) algorithms, where two problem solvers, starting from the initial and goal states, physically move toward each other. To evaluate the RTBS performance, two kinds of algorithms are proposed and are compared to real-time unidirectional search. One is called centralized RTBS where a supervisor always selects the best action from all possible moves of the two problem solvers. The other is called decoupled RTBS where no supervisor exists and the two problem solvers independently select their next moves.
Experiments on mazes and n-puzzles show that 1) in clear situations decoupled RTBS performs better, while in uncertain situations, centralized RTBS becomes more efficient, and that 2) RTBS is more efficient than real-time unidirectional search for 15- and 24-puzzles but not for randomly generated mazes. It will be shown that the selection of the problem solving organization is the selection of the problem space, which determines the baseline of the organizational efficiency; once a difficult problem space is selected, the local coordination among problem solvers hardly overcome the deficit.

We consider the case of heuristic search where the goal may change during the course of the search. For example, the goal may be a target that actively avoids the problem solver. We present a moving-target search algorithm (MTS) to solve this problem, We prove that if the average speed of the target is slower than that of the problem solver, then the problem solver is guaranteed to eventually reach the target in a connected problem space.
The original MTS algorithm was constructed with the minimum operations necessary to guarantee its completeness, and hence is not very efficient. To improve its efficiency, we introduce ideas from the area of resource-bounded planning into MTS, including
1) commitment to goals, and
2) deliberation for selecting plans.
Experimental results demonstrate that the improved MTS is 10 to 20 times more efficient than the original MTS in uncertain situations.

Recent production system applications have been experiencing exceedingly difficult software maintenance problems. This is because the control of rule firings has been buried in the production rules themselves. To cope with this problem, we propose a meta-level control architecture for production systems, where procedural programming languages, such as Lisp and C, are employed to explicitly describe the control plans of production systems. The key idea of the architecture is to view production systems as a collection of independent rule processes, each of which monitors the global database and performs actions when its conditions are satisfied by the database. Procedural Control Macros (PCMs), which are based on Hoare's CSP, are then introduced into procedural programming languages to establish communication with the collection of rule processes.
Although the PCMs are simple and easy to implement, the readability and maintainability of production system applications are greatly enhanced. Together with the original facilities of procedural languages, the PCMs enable users to efficiently specify the control plans for production systems. Furthermore, since control information is gathered into control plans, production rules can be declarative and thereby application-independent. This new feature makes it possible to develop large-scale shared rule bases. Experiments of redescribing 370 rules have been performed based on the proposed meta-level control architecture. The results demonstrate significant improvements in the readability and maintainability of those rules without any significant representation (file volumes) and runtime (processing time) overheads.

As the scale of rule-based expert systems increases, the efficiency of production systems becomes a pressing concern. Recently developed production systems thus enable users to specify an appropriate ordering or clustering of,join operations. Various efficiency heuristics have been introduced to optimize production rules manually. However, since the heuristics often conflict with each other, users have to proceed by trial and error. The problem addressed in this paper is how to automatically determine efficient join structures for production system programs. Our algorithm does not directly apply efficiency heuristics to programs, but rather enumerates possible join structures under various constraints and selects the best one. For this purpose, the cost model for production systems is introduced to estimate the run-time cost of join operations. Evaluation results demonstrate that the proposed algorithm can generate programs that are as efficient as those obtained by manual optimization, and thus can reduce the burden of manual optimization.

Organization has emerged as a key concept for structuring the activities of collections of problem-solvers. Organization self-design (OSD) has been studied as an adaptive approach to long term, strategic work-allocation and load-balancing. In this paper, we introduce two new reorganization primitives, composition and decomposition. They change the population of agents and the distribution of knowledge in an organization. To create these primitives, we formalize organizational knowledge, which represents knowledge of potential and necessary interactions among agents in an organization. We develop computational OSD techniques for agents with architectures based on production systems to take advantage of the well-understood body of theory and practice.
We first extend parallel production systems, where global control exists, into distributed production systems, where problems are solved by a society of agents using distributed control. We then introduce OSD into distributed production systems to provide adaptive work allocation. Simulation results demonstrate the, effectiveness of our approach in adapting to changing environmental demands. In addition to introducing advanced techniques for flexible OSD, our approach impacts production system design, and improves our ability to build production systems that can adapt to changing real-time constraints.

To improve the performance of forward chaining production systems, a new parallel execution model is proposed, where multiple production rules are fired in parallel in each production cycle. This paper provides solutions for three major problems in parallel rule firing. The interference analysis, which detects cases where a parallel firing result is different from the result of any sequential firing. Based on a data dependency graph of production systems, we propose general techniques applicable to both compile- and run-time interference analyses. The parallel firing algorithms, which achieve the parallel firing on multiple processor architectures. An efficient selection algorithm is provided to select multiple rules to be fired in parallel by combining the compile-and run-time interference analysis techniques. The parallel programming environment, which provides language facilities to enable programmers to make full use of potential parallelism without considering the internal parallel mechanisms. A parallel firing simulator is also implemented to estimate the effectiveness of parallel firings of production system programs. We have evaluated the effectiveness of parallel rule firings on several production system applications. Results show that the degree of concurrency can be increased by a factor of 2–9. The sources of parallelism are investigated based on the evaluation results. © 1991 IEEE

The production system is one in which mutually independent rules communicate with each other through a common database aiming at the solution of a problem. Because of its nature, the production system seems to contain a larger potential parallelism than the procedural program. At present, however, only one rule succeeded in condition matching, is selected and executed, which prevents the full utilization of the parallelism. The aim of this paper is to realize a high‐speed execution of the production system by executing the rules in parallel as far as possible. First, a new execution model is proposed for the production system, where the rules succeeded in condition match are executed in parallel as far as possible. Then the data dependency graph for the production system is introduced, and a method is presented by which more than one rule is analyzed to determine whether or not they can be executed in parallel. A method is presented which applies the result of parallel executability analysis to the parallel execution of the rules on a parallel computer, or the execution scheduling of the rules on a sequential computer. A production system PLANET was constructed which can analyze a program by a parallel execution simulation, and the proposed method was evaluated. It was verified that four to eight rules can be executed in parallel in the range of programs used in the evaluation. By combining the result of this paper with the traditional parallel condition matching, it is expected that a further parallelism can be realized. Copyright © 1989 Wiley Periodicals, Inc., A Wiley Company

（中国科学院科学出版社, 人工智能, 2003）

Lecture Notes in Artificial Intelligence 878

Interactive agents are expected to play essential roles in many fields. Moreover, it is becoming more and more important for interactive agents to deal with multilingual issues because of globalization. In this paper, we aim at realizing a multilingual interactive agent that can support end users in various fields. To achieve this goal, we first propose a framework for the multilingual interactive agent that can support dialogues with end users through Wizard of Oz. We then propose the approach of gradual acquisition of multilingual dialogues to improve the performance of the interactive agent. Moreover, we apply the proposed framework and implement a multilingual interactive agent in the field of healthcare.

The constraint-based approach has been proven useful for inducing bilingual lexicons for closely-related low-resource languages. When we want to create multiple bilingual dictionaries linking several languages, we need to consider manual creation by bilingual language experts if there are no available machine-readable dictionaries are available as input. To overcome the difficulty in planning the creation of bilingual dictionaries, the consideration of various methods and costs, plan optimization is essential. We adopt the Markov Decision Process (MDP) in formalizing plan optimization for creating bilingual dictionaries
the goal is to better predict the most feasible optimal plan with the least total cost before fully implementing the constraint-based bilingual dictionary induction framework. We define heuristics based on input language characteristics to devise a baseline plan for evaluating our MDP-based approach with total cost as an evaluation metric. The MDP-based proposal outperformed heuristic planning on total cost for all datasets examined.

While traveling, foreign visitors encounter new products that they need to understand. One solutionis by making Culturally Situated Associations (CSA) i.e. relating the products they encounter to products in their own culture. We propose the design of a system that provides tourists with CSA to help them understand foreign products. In order to provide tourists with CSA that they can understand, we must gather information about their culture, provide them with the CSA, and make sure they understand it. To deliver CSA to foreign visitors, two types of data are needed: data about the products, their associated properties and relationships, and data about the tourist cultural attributes such as country, region, language. The properties and relationships about countries, regions and products, can be extracted from open linked data on the web, and CSA can then be constructed. However, information about the tourist’s cultural attributes and the knowledge they can relate to is unavailable. One way to tackle this problem would be to extract the tourist’s cultural attributes that are needed in each situation through dialogue systems. In this case, a Culturally Situated Dialogue (CSD) must take place. To implement the dialogue, dialogue systems must follow a machine-learned dialogue strategy as previous work has shown that a machine-learned dialogue strategy outperform the handcrafted dialogue approach. We propose the design of a system that uses a reinforcement learning algorithm to learn CSD strategies that can support individual foreign tourists. Since no previous system providing CSA has been implemented, the system allows the creation of CSD strategies when no initial data or prototype exists. The method is used to generate 3 different agents that learn 3 different dialogue strategies.

In multilingual collaboration, a paucity of shared language and gaps in the language backgrounds of group members could bring about imbalanced participation, which is likely to hinder problem solving, idea generation and collaborative learning. This paper proposes a model of best balanced communication based on the Quality of Messages among participants using various languages. We describe a method for selecting the languages to be used with machine translators, and how to create the best balanced communication environment. Currently, many studies on machine translators and balancing conversations have been published, but none have attempted to balance asymmetric participation in multilingual groups. Our vision allows machine translation technologies to enhance the communication between humans with different language backgrounds in terms of balancing their participation. We conduct controlled experiments and find the proposed method successfully enables users to interact and communicate with better equality while minimizing the problems that can arise from machine translation usage.

Though websites offer a direct channel for global brands to communicate in the international market, the presence of outdated and inconsistent content can potentially create poor impressions. This study foresees the occurrence of inter- and intra-regional discrepancies in information shared via country-specific websites, for example, inconsistent content in product specifications shared with customer in the Asia-Pacific and North American regions. To deal with regional discrepancies, this study examined managerial preferences for content sharing within and beyond geographic regions in global brands and identified traits such as coupling and scales in sharing. High coupling in websites within Europe suggest that intra-regional discrepancies are more likely to occur in content shared inside the European region and high coupling in websites among Asia-Pacific, Europe, and Middle-East-Africa suggest vulnerability to inter-regional discrepancies. Preferences in sharing revealed in this study have implications in the design of consistency policy customized for specific regions.

This paper proposes a language service infrastructure for personal use called Language Mashup. It enables users to develop domain-specific multilingual applications on their personal devices by combining various kinds of language services created from the language resources provided by both academia and industry. To discuss the potential of Language Mashup, this paper introduces two key communication problems, and then our solution of a multilingual application that supports international meetings whose participants come from various countries and communicate with each other in their own languages.

Service composition is the technique of creating new services by combining several existing services. Composite services can be also combined with other composite services to form nested or hierarchical services. Given that service composition depends on the interoperability created by using common network protocols and invocation interfaces, a composite service can have an impractically large number of variations depending of the number of available services and the composite's structure. It is hard to enumerate and maintain all variations possible. To solve this problem, we introduce a higher-order function that can take functions as parameters to allow function invocation. In concrete terms, we propose the following methods: (1) a hierarchical service composition description by introducing higher-order functions and (2) a method to implement (1) in an existing composite service execution system. As a test, we apply the proposals to Language Grid, and evaluate the results. They show that our methods can reduce the number of variations that need to be registered and managed even though their overheads are quite practical.

With the proliferation of Web services, more and more functionally equivalent services are being published by service providers on the Web. Although more services mean more flexibility for consumers, it also increases the risk of choosing as consumers may have little or no past experience with the service they will interact with. Therefore, reputation systems have been proposed and are playing a crucial role in the service-oriented environment. Current reputation systems are mainly built upon the explicit feedback or rating given by consumers after experiencing the service. Unfortunately, services at the cold-start stage, prior to being rated, face the rating scarcity problem. In this paper, we focus on this problem and address it through a novel reputation model that uses the Elo algorithm to consider consumer implicit information in a graph analysis approach. Theoretical analysis is conducted to identify the sufficient and necessary condition for the model to converge to a stable state. Furthermore, experiments confirm our model outperforms the widely adopted reputation algorithm in both accuracy and convergence in the situation of rating scarcity.

Complex event processing (CEP) is attracting much attention as a method for analyzing streaming data in the IoT environment. Since a CEP system selects and executes a rule from rules that match identified events, i.e., multiple rules are sequentially executed. This, however, causes a problem in the Internet of Things (IoT) environment since rule execution is slow and the remaining rules must wait to be executed. Simply executing rules in parallel may trigger interference between rules, and thus unexpected and undesirable results
This paper extends the traditional CEP system by developing an execution model for parallel rule firing in the IoT environment so as to be able to execute multiple rules in parallel without any interference. We start with an extended parallel firing condition by adding the definition of dependency between sensors and actuators to the condition. Next, we extend synchronization control of parallel firing so as to avoid the interference among rules that can occur when actions take a lot of time to execute. This paper reveals that the extended CEP system (i) realizes triple parallelism (i. e., data parallelism, task parallelism, and pipeline parallelism) in the IoT environment and (ii) avoids the case where rule execution triggers unexpected results.

Previous studies have suggested many technologies to support non-native speaker comprehension in real-time communication. However, such technologies may impose an extra burden on non-native speakers (NNSs) if they do not match their current needs. To design a system that adapts to the changing needs of NNSs, we need to understand the types of problems NNSs face and how these problems are perceived by them. To explore such issues, we conducted a laboratory experiment with 40 NNSs (and 20 native speakers as a control group) who engaged in a listening task. During the task, the participants pressed a button whenever they encountered a comprehension problem. Next they explained each problem, the point at which they recognized the problem, and for how long it persisted. Our analysis identified twelve types of listening comprehension problems, which we further classified into three patterns based on their persistence and the time taken to perceive them. Our findings have implications for designing adaptive technologies to support listening comprehension of NNSs in real-time communication.

Internet of Things (IoT) applications are naturally dynamic in that unexpected changes and failures may occur on smart devices, on network connection and interaction between the devices, or in the environment where they are situated. To maintain the robustness of IoT applications, smart devices should be able to adapt the way they operate upon changes and failures. However, the adaptive capability of smart devices may cause IoT applications to exhibit unpredictable behavior at runtime, and thus the adaptive behaviors should be implemented in a simulated environment before they are realized in the actual environment. This paper proposes a programming language extended from AgentSpeak(L) for implementing adaptive smart devices as multiagent systems. To easily implement adaptive behavior, the programming language should allow the separation of adaptive capabilities and core functionalities. Such feature has not been supported by existing agent-oriented programming languages that tend to mix the two. The case study demonstrates that the proposed language improves maintainability of the programs compared to that of AgentSpeak(L).

At the beginning of the new millennium, we proposed the concept of intercultural collaboration where participants with different cultures and languages work together towards shared goals. Because intercultural collaboration is a new area with scarce data, it was necessary to execute parallel experiments in both in real fields as well as in research laboratories. In 2002, we conducted a one-year experiment with Japanese, Chinese, Korean and Malaysian colleagues and students to develop open-source software oriented towards machine translation. From this experiment, we understood the necessity of a language infrastructure on the Internet that could create customized multilingual environments for various situations. In 2006, we launched the Language Grid project to realize a federated operation of servers for language services. Using the Language Grid, we worked with a nongovernmental organization since 2011 to support knowledge communications between agricultural experts in Japan and farmers in Vietnam via their children. We observed that a large community emerged to utilize these nonmature machine translation technologies. During these experiences, by facing different types of difficulties, we gradually came to understand the nature of intercultural collaboration. Problems are wicked and not easily defined because of their nested and open networked origin. Fortunately, multiagent technologies can be applied to model and simulate intercultural collaboration so as to predict the difficulties and to prepare a better support systems. In this paper, we provide a brief history of the research and practice as regards intercultural collaboration and support systems.

Parallel execution and cloud technologies are the keys to speed-up service invocation when processing large-scale data. In SOA, service providers normally employ policies to limit parallel execution of the services based on arbitrary decisions. In order to attain optimal performance improvement, service users need to adapt to parallel execution policies of the services. A composite service is a combination of several atomic services provided by various providers. To use parallel execution for greater composite service efficiency we need to optimize the degree of parallelism (DOP) of the composite services by considering policies of all atomic services. We propose a model that embeds service policies into formulae to calculate composite service performance. From the calculation, we predict the optimal DOP for the composite service. Extensive experiments are conducted on real-world translation services. The results show that our proposed model has good prediction accuracy in identifying the optimal DOPs. Our model correctly predicts the optimal DOP in most cases.

In human communication, nonverbal information such as gestures and facial expressions often plays a greater role than language; and some gesture-driven operations of the latest mobile devices have proved to be easy-to-use and intuitive interfaces. In this paper we propose a method of analyzing gestures that focuses on human communication based on the dynamical kinematic model. We have extended the analysis method of our proposed approach to take into account additional effects, such as those exerted by external forces, and we analyze the effects over the entire body of forces generated by gestures. We found that the degree of exaggeration could be quantified by the value of, and changes in, torque values. Moreover, when calculating them taking into account external forces and the moment of drag that is acting on both feet, it is possible to determine the twisting torque of the main joints with a high degree of precision. We also noted "preparation" or "follow-through" motions just before and after the emphasized motion, and found that each behavior can be quantified by an "undershoot" or "overshoot" value of changes in torque.

This paper reports an application for smartphones that allows users to use digits for simple keyword input and its usability evaluation. This input method, called “one touch character”, helps users to enter keywords into a search engine entry field. Its easy-to-learn and easy-to-use characteristics make it highly effective in encouraging the elderly, who generally dislike IT systems, to search for information on the Internet. This smartphone application was tested by 30 subjects, ranging from young adults to the elderly, from the aspects of operation time and error rate. “One-touch-character” is also compared with other character input methods, such as “multi-touch-per-character” and “flick”.

知識の専門家と利用者の言語が異なる多言語知識コミュニケーションにおいては,コミュニケーションモデルを反復的に改善するプロセスが重要だが,このモデルを明確に分析・評価する方法はない.本研究では,計算機介在のコラボレーションにおいて重視されるアウェアネスの概念を取り入れ,ベトナム農業支援プロジェクトにおける,日本人専門家からベトナム人児童に対する知識提供を例題として,双方間でなされた質問応答文とインタビュー調査を分析し相互の状況認識差を明らかにする.その結果,(i)約4割の質問応答文でアウェアネスの欠落があること,(ii)アウェアネスが欠落していても殆どの質問応答文が知識伝達に負の影響がないことを明らかにした.

The Language Grid is a service-oriented collective intelligence platform, which provides a multilingual environment to support intercultural collaboration activities by allowing users to create language services from existing language resources for their own requirements. To protect intellectual property right of language resources, the existing Language Grid has employed a strict agreement which does not allow personal and commercial use. This agreement conflicts with some language resources' policies which are provided for commercial use and which are limited to personal use. To remove these limitations we introduced a personalized language grid platform called Language Mashup. This platform enables users to operate a language grid on user's computer for his/her personal use of language resources. However, this enforces user to create wrappers of language resources and register to Language Mashup by themselves, which is difficult and time consuming task. To realize Language Mashup, in this paper we describe how wrappers can be deployed and registered in different personal environments, and how to reduce the cost of creating wrappers by introducing deployment manager and wrapper repository. We also introduce an ecosystem of Language Grid which including different types of Language Grid to share language resources with different policies.

クラウドソーシングを複雑なタスクに適用する試みが行われている.その多くは,タスクをサブタスクに分解し,それらをワークフローによって制御するものである.しかしクラウドソーシングでは,サブタスクを実行する作業者の能力や信頼性が未知であるため,最適なワークフローを事前に構成することができない.そこで,作業者が柔軟に協調できるように,黒板アーキテクチャをクラウドソーシングの制御に適用する.例えば,複数のサブタスクの処理を複数の作業者に並行して依頼し,実行結果に基づき次に処理すべきサブタスクを決定する.このように黒板アーキテクチャのデータ駆動モデルにより,サブタスクの処理結果の信頼性に応じた動的な制御を実現する.また,サブタスクの実行結果の記録に基づいて作業者の能力を推定し,それ以降の実行制御に役立てる.提案手法を複雑なタスクである日尼翻訳に適用し,黒板アーキテクチャが如何に効果的に働くかを示す.

High quality bilingual dictionaries are rarely available for lower-density language pairs, especially for those that are closely related. Using a third language as a pivot to link two other languages is a wellknown solution, and usually requires only two input bilingual dictionaries to automatically induce the new one. This approach, however, produces many incorrect translation pairs because the dictionary entries are normally are not transitive due to polysemy and the ambiguous words in the pivot language. Utilizing the complete structures of the input bilingual dictionaries positively influences the result since dropped meanings can be countered. Moreover, an additional input dictionary may provide more complete information for calculating the semantic distance between word senses which is key to suppressing wrong sense matches. This paper proposes an extended constraint optimization model to inducing new dictionaries of closely related languages from multiple input dictionaries, and its formalization based on Integer Linear Programming. Evaluations indicated that the proposal not only outperforms the baseline method, but also shows improvements in performance and scalability as more dictionaries are utilized.

Service composition involves combining many existing services to solve a complex task in the service-oriented environment. Due to the progress in modern smart devices, mobile devices can now act as service providers. The mobility of mobile service providers and the dynamism of the mobile environment are unique features. As a result, there are two issues with service composition in mobile environments: (1) how to use the mobility of devices as an non-functional Quality of Service (QoS) criterion, and (2) how to deal with the dynamism of the mobile environment to realize the optimal service composition. In order to deal with above issues, we proposed a QoS-aware service composition model for service composition in mobile environments considering the feature of mobility in the mobile environment. Moreover, we proposed an adaptive approach called k-neighbor. The main idea of k-neighbor approach is to decompose the composite service into smaller elementary composite services then to set up a solution plan for those elementary composite services based on the available provider set. The evaluation shows that the way of decomposing a composite service adapts to different dynamism of the environment.

The recent popularity of machine translation has increased the demand for the evaluation of translations. However, the traditional evaluation approach, manual checking by a bilingual professional, is too expensive and too slow. In this study, we confirm the feasibility of crowdsourcing by analyzing the accuracy of crowdsourcing translation evaluations. We compare crowdsourcing scores to professional scores with regard to three metrics: translation-score, sentence-score, and system-score. A Chinese to English translation evaluation task was designed using around the NTCIR-9 PATENT parallel corpus with the goal being 5-range evaluations of adequacy and fluency. The experiment shows that the average score of crowdsource workers well matches professional evaluation results. The system-score comparison strongly indicates that crowdsourcing can be used to find the best translation system given the input of 10 source sentence.

Integrating language resources and language services is a critical part of building natural language processing applications. Service workflow and processing pipeline are two approaches for sharing and combining language resources. Workflow languages focus on expressive power of the languages to describe variety of workflow patterns to meet users' needs. Users can combine those language services in service workflows to meet their requirements. The workflows can be accessible in distributed manner and can be invoked independently of the platforms. However, workflow languages lack of pipelined execution support to improve performance of workflows. Whereas, the processing pipeline provides a straightforward way to create a sequence of linguistic processing to analyze large amounts of text data. It focuses on using pipelined execution and parallel execution to improve throughput of pipelines. However, the resulting pipelines are standalone applications, i.e., software tools that are accessible only via local machine and that can only be run with the processing pipeline platforms. In this paper we propose an integration framework of the two approaches so that each offests the disadvantages of the other. We then present a case study wherein two representative frameworks, the Language Grid and UIMA, are integrated.

Bilingual dictionaries are vital in many areas of natural language processing, but such resources are rarely available for lower-density language pairs, especially for those that are closely related. Pivot-based induction consists of using a third language to bridge a language pair. As an approach to create new dictionaries, it can generate wrong translations due to polysemy and ambiguous words. In this paper we propose a constraint approach to pivot-based dictionary induction for the case of two closely related languages. In order to take into account the word senses, we use an approach based on semantic distances, in which possibly missing translations are considered, and instance of induction is encoded as an optimization problem to generate new dictionary. Evaluations show that the proposal achieves 83.7% accuracy and approximately 70.5% recall, thus outperforming the baseline pivot-based method.

Machine translation is increasingly used to support multilingual communication. Because of unavoidable translation errors, multilingual communication cannot accurately transfer information. We propose to shift from the transparentchannel metaphor to the human-interpreter (agent) metaphor. Instead of viewing machine translation mediated communication as a transparent channel, the interpreter (agent) encourages the dialog participants to collaborate, as their interactivity will be helpful in reducing the number of translation errors, the noise of the channel. We examine the translation issues raised by multilingual communication, and analyze the impact of interactivity on the elimination of translation errors. We propose an implementation of the agent metaphor, which promotes interactivity between dialog participants and the machine translator. We design the architecture of our agent, analyze the interaction process, describe decision support and autonomous behavior, and provide an example of repair strategy preparation. We conduct an English-Chinese communication task experiment on tangram arrangement. The experiment shows that, compared to the transparent-channel metaphor, our agent metaphor reduced human communication effort by 21.6%. Copyright © 2013 ACM.

The availability of documents in multiple languages provides a challenge to maintain consistency of information in the multilingual documents. In this research we present a mechanism to manage inconsistencies in the parallel multilingual documents. We propose a state transition model to define the states of the sentence in the multilingual documents, the set of actions performed on these sentences, the set of transition functions that describe the state transition of the sentences when actions are performed on them. We then define consistency checking with a set of consistency rules to signal states of the sentences leading to inconsistencies. We apply our mechanism for managing inconsistencies in the multilingual document on "Traffic Safety Guidelines for Pedestrians and Cyclists" issued by Japan Traffic Safety Association by defining the state transition model of the multilingual contents and applying consistency checking and handling for managing inconsistencies. We found that our mechanism greatly improves the task of locating inconsistent portions in the document and the inconsistent versions of the multilingual documents and guaranteeing the consistency of information in multilingual documents.

異言語間において知識を伝達することは国際的な活動にとって非常に重要な課題である.本研究では,ベトナムの児童を仲介して日本人の専門家がベトナムの農業従事者に知識を伝達する,YMC-Vietプロジェクトをケーススタディとして取り上げる.本稿では,日本人専門家とベトナム人児童間の多言語知識コミュニケーションを分析し,生じていたミスコミュニケーションを明らかにした.知識を一方的に伝達するだけのコミュニケーションモデルであるために問題が生じていたため,相互的なコミュニケーションを実現できるようなコミュニケーションパスを導入することを提案する.

YMC-Vietというプロジェクトでは,日本人専門家の知識を翻訳技術を用いてベトナム人稲作農家へ伝達する事業を行っている.日本語からベトナム語へ翻訳を行う場合,機械翻訳を適用するだけでは翻訳精度が十分ではない,人手による翻訳は,機械翻訳よりも高い精度の翻訳ができるがコストが高い.本研究では,機械翻訳と翻訳前編集を組み合わせた翻訳プロセスを検討する.一般の人間が担当可能な前編集により,機械翻訳の品質を改善することが狙いである.

YMC-Vietというプロジェクトでは，日本人専門家の知識を翻訳技術を用いてベトナム人稲作農家へ伝達する事業を行っている．日本語からベトナム語へ翻訳を行う場合，機械翻訳を適用するだけでは翻訳精度が十分ではない，人手による翻訳は，機械翻訳よりも高い精度の翻訳ができるがコストが高い．本研究では，機械翻訳と翻訳前編集を組み合わせた翻訳プロセスを検討する．一般の人間が担当可能な前編集により，機械翻訳の品質を改善することが狙いである．

In the machine translation mediated communication, miscommunication emerges because of the ignorable translation errors It is proposed to shift from the transparent-channel metaphor for machine translator to the human-interpreter (agent) metaphor Interactivity between agent and communication participants will help eliminate the translation errors, the noise of the channel However, there is no unique model to describe these interactions We propose a simple conceptual model to describe the interactivity We analyzed concrete interactions, and described a three-interaction model Based on it, we clarified three interactivity levels to category the agent's ability to eliminate translation error and clear the miscommunication After examining several existing works of interactions, we argued that our proposal will be helpful for better understanding of interactivity for machine translation mediated communication

With the development of service-oriented computing environments, more and more Web services have become available. Service providers have to consider how to design composite services to meet requirements from various users. However, QoS of composite service is difficult to evaluate and predict due to the uncertainty and multiple metrics of QoS, which makes it necessary for user-centered design of composite services. Moreover, combining human services and Web services is becoming an important issue in service composition in cases that Web services cannot meet the requirements from users. Therefore, it is necessary to test the service composition environments for human-computer interaction and study human behaviors. In this paper, we address the above issues by proposing a participatory service design approach based on user-centered QoS considering a real-world case of field-based multi-language communication service design. To achieve this goal, we first describe the QoS model for service design considering users' requirements. Then, we propose the participatory service design process, consisting of service model refinement, participatory simulation, QoS evaluation and QoS data update. Finally, we use the field study of multi-language service design for Vietnamese agricultural knowledge communication to illustrate our proposed design methodology. © 2013 IEEE.

The future of computing systems is leading to a growing complexity of heterogeneous computing systems integration situated in highly dynamic and uncertain environment. Self-organization has been proposed to be one of the solutions due to its autonomous capability to govern itself and to adapt upon changes. Developing self-organizing systems requires a suitable modeling method because of its unique characteristics, such as self-adaptive, decentralized, and emergent. Existing multiagent systems engineering methodologies lack modeling methods to design behavior adaptation for self-organizing systems. This paper proposes a role modeling method that allows designers to model how agents perform behavior adaptation at runtime by making transitions from one set of roles to another in decentralized manner. The behavior adaptation of each individual agent may eventually emerge as the adaptation of the organization. The proposed modeling method also helps designers to analyze issues that may occur because of playing multiple roles, such as role conflict, role overload, role transition oscillation, and other common issues in concurrency. © 2013 IEEE.

In this paper, we focused on the characteristic gestures which producing the sense of realism and intimacy in web communication using 3DAgent, and proposed the gesture exaggeration mechanical model which representing the "reservoir (Tame)" or "deciding actions (Kime)" effectively in Anime and Kabuki. By analyzing some tens of gestures including the natural motion and exaggerating one which had same start and end positions, we obtained the following results. The degree of exaggeration of behavior can be expressed mechanically by the integral value of the joint torque. The ratio of the torque value GER between the portion containing the exaggerated behavior and other position has value within the certain range from 0.4 to 0.6. This value can be used as an indicator in determining the good balanced gesture exaggeration. Moreover, we found that the result could also be seen in the case of exaggeration gestures used effectively such as the speech and presentation of famous persons, so we could determine the gesture exaggeration by GER. © 2013 Springer-Verlag Berlin Heidelberg.

High quality machine readable dictionaries are very useful, but such resources are rarely available for lower-density language pairs, especially for those that are closely related. In this paper, we proposed a heuristic framework that aims at inducing one-to-one mapping dictionary of a closely related language pair from available dictionaries where a distant language is involved. The key insight of the framework is the ability to create heuristics by using distant language as pivot, incorporate given heuristics, and an iterative induction mechanism that human interaction can be potentially integrated. An experiment based on basic heuristics regarding syntactics and semantics resulted in up to 85.2% correctness in target dictionary with correctness of major part reached 95.3%, which proved that we can perform automated creation of a high quality dictionary with our framework. © 2013 IEEE.

Translation practices in parallel multilingual documents are bound to generate inconsistencies in the documents due to missing information or part of document not translated, changes not propagated across languages, unavailability of same information in all languages. In this research, we present a mechanism to manage inconsistencies in parallel multilingual documents. To model inconsistencies in the multilingual contents, we propose a state transition model to define the states of sentences in the documents, the actions performed on the sentences and the set of transition functions. We then define consistency rules to signal the states of sentences resulting in inconsistencies. We illustrated the proposed mechanism with a case study on tracking inconsistencies in the multilingual MediaWiki Installation Guide as a tool support for tracking inconsistent portions such as missing information, changes not propagated and unavailability of same information in multilingual documents.

Previous research has shown that audio communication is particularly difficult for non-native speakers (NNS) during multilingual collaborations. Especially when audio signals become distorted, NNS are overburdened by not only having to communicate with imperfect language skills, but also compensating for the deteriorations. Under these faulty audio conditions, NNS need to pay extra time and effort to understand the conversation. In order to give NNS more time to process conversations, we tested the insertion of silent gaps (from 0.2 to 0.4 seconds) between conversational turns. First, gaps were inserted into a previously taped conversation, resulting in a significant improvement of NNS's understanding of the conversation. Second, gaps were inserted during a real-time audio conference by adding artificial delay between native speakers. The results show that the added delays have a combination of beneficial and detrimental effects for both native and non-native speakers. The findings have implications towards how audio conferencing can be improved for NNS. Copyright 2013 ACM.

The emergence of crowdsourcing makes various types of tasks done in more and more efficient ways with low costs. Meanwhile, with the development of service-oriented computing, more and more Web services are provided for users. However, there are situations that Web services cannot meet users' QoS requirements. Therefore, pure Web services are always not enough when constructing business processes. In this paper, we propose the composition of crowdsourced services and Web services for QoS improvement. Further, this study aims at analyzing how QoS dimensions of crowdsourced services and Web services affect each other in the business processes. We use an example of the translation process to illustrate the proposed approach.

Here we use the term "field" to refer to a sphere of practical operation, and correspondingly the term "field informatics" describes informatics tools and methodologies that arise in the field. The components of field informatics are description, prediction, design and transfer, and the methods for those components vary widely. For example, we consider the social goal of revitalizing a mountainous area experiencing depopulation and we show how the tools and methodologies of field informatics may be used to describe such situations using remote sensing, biologging, human sensing and ethnography
the effects of various solutions can be predicted using system dynamics and multiagent simulations
the solutions can be designed using inclusive design or participatory design methods
and finally the experience gained can be transferred using case writing and outreach communication.

With the development of service-oriented computing, more and more Web services are provided for users. However, there are situations that services or service processes cannot meet users' requirements in functional QoS dimensions (e.g., translation quality in a machine translation service). Meanwhile, the emergence of crowdsourcing makes various types of tasks done in more and more efficiency ways with low costs. To consider both functional QoS and non-functional QoS, in this paper we try to combine crowdsourcing activities with service processes. Further, this study aims at analyzing the effects of crowd activities on service processes in the real world. We use a case study in the domain of language service with a large scale experiment to show that composing crowd activities and Web services brings variety and creativity to the traditional service processes and human processes. From the experiments and analysis, we find out that quality of crowd activities is essential to service processes, and that crowd activities with high quality can significantly improve various QoS dimensions of the service processes. © 2012 Springer-Verlag.

In recent years, collaborative translation has become more and more important for translation volunteers to share knowledge among different languages, among which Wikipedia translation activity is a typical example. During the collaborative translation processes, users with different mother tongues always conduct frequent discussions about certain words or expressions to understand the content of original article and to decide the correct translation. To support such kind of multilingual discussions, we propose an approach to embedding a service-oriented multilingual infrastructure with discussion functions in collaborative translation systems, where discussions can be automatically translated into different languages with machine translators, dictionaries, and so on. Moreover, we propose a Meta Translation Algorithm to adapt the features of discussions for collaborative translation, where discussion articles always consist of expressions in different languages. Further, we implement the proposed approach on LiquidThreads, a BBS on Wikipedia, and apply it for multilingual discussion for Wikipedia translation to verify the effectiveness of this research. © 2012 IEEE.

With the development of service-oriented computing environments, QoS-aware service selection has been a more and more important research issue. In service composition environments, QoS attributes of atomic services are always aggregated for computing the QoS of the composite services, which has been reported in many previous studies. However, there are situations that some QoS attributes cannot be aggregated for composite services. For example, it is difficult to compute the translation quality of a composite translation service by simply aggregating its component atomic services (machine translation service, morphological analysis service, dictionary service). Moreover, when multiple QoS attributes are used for evaluating services, it is always difficult to maximize all the QoS attributes because there might be anti-correlated relations between them. To address above problems, this paper proposes an approach for selecting services based on context-aware factors of QoS attributes. In our proposed approach, context-aware factors that affect QoS attributes are first extracted from analyzing their correlation with QoS attributes. Then, QoS data is generated based on the extracted factors for QoS prediction and evaluation. Further, dynamic service selection is realized based on QoS prediction and evaluation considering user requirements. We use a case study in the domain of language service with some experiments to show the effectiveness of our approach. © 2012 IEEE.

QoS computation plays an important role in Web service selection. It involves property value preprocessing aspect, user satisfaction calculation aspect, and aggregation of multiple QoS properties aspect. However, little attention has been paid to users participating in QoS computation. In this paper, we examine QoS computation from the angle of experienced users and novice users. An experienced user is able to be more active in providing configuration information such as expected boundary of a QoS property, distribution function of user satisfaction, and the aggregation weight of each QoS property. While a novice user has limited experience to do this. Based on the study of user-centered factors in QoS computation, we propose a user-centered QoS computation, which provides a new choice of normalization in property value preprocessing aspect, an approach of approximation in user satisfaction calculation aspect, and a weight suggestion way in aggregation of multiple properties aspect. A case study in translation service selection shows that the proposed user-centered QoS calculation is more efficient for novice users than random configuration, and much more efficient for experience users. © 2012 IEEE.

Service grid is an infrastructure for service-oriented collective intelligence. It provides a set of enabling functionalities to support coordination of services, such as service registries, service composition, access control, and monitoring. To form the service-oriented collective intelligence, various types of services need to be connected on the service grid, and managed by the service grid operator. However, it is difficult for single service grid operator to gather and organize services in various domains. Therefore, building service grids in different domains and connecting these service grids are essential for expanding service-oriented collective intelligence across domains. To this end, we have designed a service domain model to specialize general-purpose service grid to a specific domain and realize interoperability among service grids. Moreover, we have also developed service grid federation architecture to share service registries, compose services across service grids, and control and monitor accesses to the composite services. Finally, we have applied the proposed architecture to the language service domain to construct the Language Grid. © 2012 IEEE.

An increased number of machine translation services are now available. Unfortunately, none of them can provide adequate translation quality for all input sources. This forces the user to select from among the services according to his needs. However, it is tedious and time consuming to perform this manual selection. Our solution, proposed here, is an automatic mechanism that can select the most appropriate machine translation service. Although evaluation methods are available, such as BLEU, NIST, WER, etc., their evaluation results are not unanimous regardless of the translation sources. We proposed a two-phase architecture for selecting translation services. The first phase uses a data-driven classification to allow the most appropriate evaluation method to be selected according to each translation source. The second phase selects the most appropriate machine translation result by the selected evaluation method. We describe the architecture, detail the algorithm, and construct a prototype. Tests show that the proposal yields better translation quality than employing just one machine translation service.

Due to instant availability and low cost, machine translation is becoming popular. Machine translation mediated communication plays a more and more important role in international collaboration. However, machine translators cannot guarantee high quality translation. In a multilingual communication task, many in-domain resources, for example domain dictionaries, are needed to promote translation quality. This raises the problem of how to help communication task designers provide higher quality translation systems, systems that can take advantage of various in-domain resources. The Language Grid, a service-oriented collective intelligent platform, allows in-domain resources to be wrapped into language services. For task-oriented translation, we propose service composition scenarios for the composition of different language services, where various in-domain resources are utilized effectively. We design the architecture, provide a script language as the interface for the task designer, which is easy for describing the composition scenario, and make a case study of a Japanese-English campus orientation task. Based on the case study, we analyze the increase in translation quality possible and the usage of in-domain resources. The results demonstrate a clear improvement in translation accuracy when the in-domain resources are used.

Since various communities withmultiple languagesinteract in daily life, tools are needed to support intercultural communication. However, we often observe that the success of a multi-language tool in one situation does not guarantee its success in another. To develop multi-language environments that can handle various situations in various communities, existing language resources should be easy to share and customize. Therefore, we designed the Language Grid as service-oriented collective intelligence; it allows users to freely create language services from existing language resources and combine them to develop new services to meet their own requirements. This paper explains the design concept and service architecture of the Language Grid, and our approach to user involvement in collective intelligence activities. An institutional design is also essential forcollective intelligence. We create a federated operation model to bridgedifferentstake-holders including service providers, service users, and service grid operators.

非母語で多人数音声会議に参加することは,大変な困難を伴う.対面会議ではほぼ問題なく会議に参加し議論できていても,多人数音声会議になると,たちまち理解度が低減し,発言権を得ることが困難になってしまう人が大半である.このような状況が続くと,非母語音の多人数音声会議における存在感が急激に低下してしまう.本研究では,電話会議中に非母語音が複数の高度な情報処理を並列に行わなければならない過負荷な状態に陥っていることに注目し,非母語音に付加的なリソース(処理時間)を与えることを試みる.

In this paper, we assumed that the nonverbal communication by using 3DAgents with natural gestures had various advantages compared with only the traditional voice and video communication, and we developed the IMoTS (Interactive Motion Tracking System) to verify this hypothesis. The features of this system are that the natural gestures of 3DAgents are captured easily by using interactive GUI from the 2D video images in which some characteristic human behaviors are captured, transmitted, and reproduced by natural gestures of 3DAgents. From the experimental results, we showed that the accuracy of captured gestures which often used in web communications was within the level of detectable limit. And we found that human behaviors could be characterized by the mathematical formula, and some of the information could be transmitted, especially some personalities such as quirks and likeness had the predominant effects of impressions and memories of human.

Quality of Service (QoS) can be used to select desired services from among those offering the equivalent function. In language services such as machine translation, one of the QoS metrics is translation accuracy. However, the problems are that evaluating the translation accuracy is too expensive, that the translation accuracy varies with the difficulty of the task, and that the usefulness of the translation to the user depends on the abilities of the user. In this paper, we propose a framework that selects a useful service for a specific user and task by using reputation information of users, which can be obtained at low cost. First, hypothetical reasoning is used to estimate the partial order relation between the accuracy of the language services, the language ability of the users, and the difficulty of the tasks. Second, deductive reasoning is applied to recommend useful services given the user and the task. We propose a reputation-based language service selection system that combines a partial order acquisition system with a service selection system. © 2011 IEEE.

The lack of accuracy is the main issue in language service research. One reason is the negligence of accommodating users and their cultures in discovering, composing, and organizing services. On the other hand, a community behavior is mostly mirrored in jargons, common sentences, mottos and so forth. Moreover, culture is inseparable to its contents. Most culture disputes are due to cultural contents that can be represented in images, audio and videos. These two features of culture analysis and contents are currently not available in language services. Our main contribution is that we provided a necessary framework of service discovery, composition and organization for cultural language service to solve this problem. In service discovery, we implemented semantic similarity in cultural language services by utilizing language service ontology. In service composition, we presented hybrid language service architecture to provide multi level analysis in cultural text and culture contents in multimedia. In service organization, we provided multi-objective constraint optimization that is able to accommodate the characteristic of language services organized in multi culture environment. © 2011 IEEE.

Consistent word selection in machine translation is currently realized by resolving word sense ambiguity through the context of a single sentence or neighboring sentences. However, consistent word selection over the whole article has yet to be achieved. Consistency over the whole article is extremely important when applying machine translation to collectively developed documents like Wikipedia. In this paper, we propose to consider constraints between words in the whole article based on their semantic relatedness and contextual distance. The proposed method is successfully implemented in both statistical and rule-based translators. We evaluate those systems by translating 100 articles in the English Wikipedia into Japanese. The results show that the ratio of appropriate word selection for common nouns increased to around 75% with our method, while it was around 55% without our method.

In current Wikipedia translation activities, most translation tasks are performed by bilingual speakers who have high language skills and specialized knowledge of the articles. Unfortunately, compared to the large amount of Wikipedia articles, the number of such qualified translators is very small. Thus the success of Wikipedia translation activities hinges on the contributions from non-bilingual speakers. In this paper, we report on a study investigating the effects of introducing a machine translation mediated BBS that enables monolinguals to collaboratively translate Wikipedia articles using their mother tongues. From our experiment using this system, we found out that users made high use of the system and communicated actively across different languages. Furthermore, most of such multilingual discussions seemed to be successful in transferring knowledge between different languages. Such success appeared to be made possible by a distinctive communication pattern which emerged as the users tried to avoid misunderstandings from machine translation errors. These findings suggest that there is a fair chance of non-bilingual speakers being capable of effectively contributing to Wikipedia translation activities with the assistance of machine translation. © 2011 IEEE.

Content providers want to make recommendations across multiple interrelated domains such as music and movies. However, existing collaborative filtering methods fail to accurately identify items that may be interesting to the user but that lie in domains that the user has not accessed before. This is mainly because of the paucity of user transactions across multiple item domains. Our method is based on the observation that users who share similar items or who share social connections, can provide recommendation chains (sequences of transitively associated edges) to items in other domains. It first builds domain-specific-usergraphs (DSUGs) whose nodes, users, are linked by weighted edges that reflect user similarity. It then connects the DSUGs via the users who rated items in several domains or via the users who share social connections, to create a cross-domain-user graph (CDUG). It performs Random Walk with Restarts on the CDUG to extract user nodes that are related to the starting user node on the CDUG even though they are not present in the DSUG of the starting user node. It then adds items possessed by those users to the recommendations of the starting node user. Furthermore, to extract many more user nodes, we employ a taxonomy-based similarity measure that states that users are similar if they share the same items and/or same classes. Thus we can set many suitable routes from the starting user node to other user nodes in the CDUG. An evaluation using rating datasets in two interrelated domains and social connection histories of users as extracted from a blog portal, indicates that our method identifies potentially interesting items in other domains with higher accuracy than is possible with existing CF methods.

Service-oriented collective intelligence, which creates new value by combining services provided by various organizations via services computing technologies, has been gaining in importance with the development of services computing technologies. Because collective intelligence needs many participants, it is crucial to build a framework where a wide variety of policies of service providers are satisfied. In this paper, we propose an architecture which handles a comprehensive process of service selection, adaptation, and coordination to satisfy policies of service providers. First the system selects services, and then adapts the services to the given policies if any of available services cannot satisfy the policies. To achieve this, we formalized this problem as an extension of constraint satisfaction problem and showed a solution. Moreover, the system often needs to force a composite service to follow protocols given by service providers. Therefore we proposed a method which uses meta-level control functions for composite services in order to change order of service execution. © 2010 IEEE.

Web services are increasingly being used in the natural language processing community as a way to increase the interoperability amongst language resources. This paper extends our previous work on integrating two different platforms, i.e. Heart of Gold and Language Grid. The Language Grid is an infrastructure built on top of the Internet to provide distributed language services. Heart of Gold is known as middleware architecture for integrating deep and shallow natural language processing components. The new feature of the integrated architecture is the combination of composite language services in the Language Grid and the multiple linguistic processing components in Heart of Gold to provide a better quality of language resources available on the Web. Thus, language resources with different characteristics can be combined based on the concept of service oriented computing with different treatment for each combination. Having Heart of Gold fully integrated in the Language Grid environment would contribute to the heterogeneity of language services.

As the number of language resources accessible on the Internet increases, many efforts have been made for combining language resources and language processing tools to create new services. However, existing language resource coordination frameworks cannot manage issues of intellectual property associated with language resources, which make it difficult for most end-users to get supports for their intercultural collaborations because they always have to deal with the issues by themselves. In this paper, we aim at constructing a new language service management architecture on the Language Grid, which enables language resource providers to control access to their resources in accordance with their own policies. Furthermore, we apply the proposed architecture to the operating Language Grid in order to validate the effectiveness of the architecture. As a result, several service management models utilizing the monitoring and access constraints are occurring to satisfy various requirements from language resource providers. These models can handle paid-for language resources as well as free language resources. Finally, we discuss further challenging issues of combining language resources under each different policies.

With the development of the Internet environments, more and more language services become accessible for common people. However, the gap between human translators and machine translators remains huge especially for the domain of localization processes that requires high translation quality. Although efforts of combining human and machine translators for supporting multilingual communication have been reported in previous research, how to apply such approaches for improving localization processes are rarely discussed. In this paper, we aim at improving localization processes by composing human and machine translation services based on the Language Grid, which is a language service platform that we have developed. Further, we conduct experiments to compare the translation quality and translation cost using several translation processes, including absolute machine translation processes, absolute human translation processes and translation processes by human and machine translation services. The experiment results show that composing monolingual roles and dictionary services improves the translation quality of machine translators, and that collaboration of human and machine translators is possible to reduce the cost comparing with the absolute bilingual human translation. We also discuss the generality of the experimental results and further challenging issues of the proposed localization processes.

With the development of services computing technology, more and more voluntary services have been available on the Internet. When using voluntary services, users tend to demand higher QoS (e.g., throughput of the services) than they actually need because there is no cost. To control QoS of the voluntary services appropriately, it is necessary to design resource allocation mechanism using utilities on both service users and providers. Therefore, we have proposed market-oriented resource allocation where users and providers exchange system resources and QoS based on their utilities. In our proposed approach, service users obtain more utilities if higher QoS is allocated according to their preferences in using the services, while service providers get more utilities if their services are more effectively used by their preferred users. In order to validate the proposed method, we have compared market-based approach with demand-based approach by simulation. The simulation results show that our approach motivated users to give true demands more than demand-based approach. © 2010 IEEE.

In this paper; we present a protocol for collaborative translation; where two non-bilingual people who use different; languages collaborate to perform the task of translation using machine translation (MT) services. Members in one real life example of intercultural collaboration try to share information more effectively by modifying unnatural machine translated sentences manually and improving their fluency. However, there are two problems with this method: One is that poor quality of translation can induce misinterpretations, and the other is that phrases in the machine translated sentence that a person cannot make sense of remain unmodified. The proposed protocol is designed to solve these problems. More concretely, one person; who handles the source language and knows the original sentence (source language side); evaluates the adequacy between the original sentence and the translation of the sentence modified to be fluent by the other person, who handles the target language (target language side). In addition; by determining whether the meaning of the machine translated sentence is understandable; it is ensured that the two non-bilingual people do above tasks properly. As a result; this protocol I) improves MT quality; and 2) terminates successfully only when the translation result becomes adequate and fluent. The experiment results show that when the protocol terminates successfully; the quality of the translation increases to about 83 percent in Japanese-English translation and 91 percent in Japanese-Chinese translation.

Service-Oriented Architectures (SOA) provide infrastructures to make resources available to other participants in the network as independent services. However, service providers, not having the autonomy to decide who they collaborate with. might be reluctant to participate in such open systems; the client being the sole responsible of the selection of services for the composition. Multiagent systems research offer sonic solutions in term of trust and reputation mechanisms as well as in coalition formation theory. This paper presents a multiagent based negotiation model to enable provider autonomy in composite web services. QoS-based reputation is built from both feedbacks retrieved from execution awl from subjective feedbacks given by the client. This model is illustrated by an example based on the Language Grid Project, an service infrastructure for language resources.

Network games are attracting attention as simulation platforms For social experiments because of their rich visualization performance and scalability. Our objective in this study is to develop a. participatory simulation platform on a network game. Unlike non player characters (NPCs) in network games, agents in a participatory multiagent-based simulation (PMAS) should behave as real-world humans according to behavior models. We developed a novel networked participatory simulation platform called, gumonji/Q by integrating scenario description language Q with the network game gumonji. This paper details the implementation of gumonji/Q. In order to connect Q and gumonji. we implement communication sub-components that realize TCP/IP communication between them, and a scenario translator to convert a request from Q into a sequence of operators. This makes it possible for the gumonji simulator to deal with human-controlled avatars and Q-controlled agents in a unified way.

The success of the emerging service oriented computing relies fully on the Quality of Service (QoS). However, existing QoS techniques do not accommodate users&apos; skills and preferences. We propose user-centered QoS, which is a QoS defined by the interaction between skills/preferences of service user(s) and quality of service provider(s). By implementing user-centered QoS approach, the best service is delivered to users based on the calculation not only the quality of the services but also the skill/information of users. We proposed a novel two-stage approach for combining services in user-centered QoS, i.e. intra-workflow and inter-workflow service selection. Intra-workflow service selection is used to calculate the most optimal QoS value for each composite service. Inter-workflow service selection is used to search for the most optimal combination of composite services by utilizing the QoS values obtained from intra-workflow service selection. In this paper, we provide a concrete example of user-centered QoS in the language services domain. This problem arises when there are multi users with different quality of English using multilingual chat service.

Machine translation services available on the Web are becoming increasingly popular. However, a pivot translation service is required to realize translations between non-English languages by cascading different translation services via English. As a result, the meaning of words often drifts due to the inconsistency, asymmetry and intransitivity of word selections among translation services. In this paper, we propose context-based coordination to maintain the consistency of word meanings during pivot translation services. First, we propose a method to automatically generate multilingual equivalent terms based on bilingual dictionaries and use generated terms to propagate context among combined translation services. Second, we show a multiagent architecture as one way of implementation, wherein a coordinator agent gathers and propagates context from/to a translation agent. We generated trilingual equivalent noun terms and implemented a Japanese-to-German-and-back translation, cascading into four translation services. The evaluation results showed that the generated terms can cover over 58% of all nouns. The translation quality was improved by 40% for all sentences, and the quality rating for all sentences increased by an average of 0.47 points on a five-point scale. These results indicate that we can realize consistent pivot translation services through context-based coordination based on existing services.

A composite Web service designed based on abstract Web services, which define only interfaces, allows an application developer to select services required for his application only by setting endpoints for the atomic Web services. In open environment, however the composite Web service configured in this manner may fail due to unique behaviors of the selected services. It is difficult for the designer of the composite Web service to prevent the failure because he does not know which services are selected and how they behave. On the other hand, the application developer is not authorized to modify the composite Web service due to the need to protect intellectual rights. Our solution is Service Supervision, which monitors and controls execution of composite Web services. Service Supervision makes the followings possible. 1) An application developer can control the behavior of a composite Web service by changing the execution state, even if the he is not authorized to modify the composite Web services. 2) A control pattern for coordinating Web services can be applied to various composite Web services in order to reduce the load imposed by designing control processes. In order to realize Service Supervision, we introduce meta-level control of a composite Web service. Moreover we then use the choreography to define the interaction protocols for the controls. The proposed framework is based on existing standard languages, WS-BPEL and WS-CDL. Therefore we can exploit existing tools and expertise of SOA engineers.

Machine translation services available on the Web are getting increasingly popular. Multiple translation services are often combined for intercultural communication. Furthermore, since English is most frequently used as a hub language, it is often necessary to cascade different translation services to realize translations between non-English languages. As a result, the word sense is often changed because of the inconsistency, asymmetry, and intransitivity of word selections between machine translation services. Therefore, we propose a context-based coordination framework in which the context is propagated among cascaded translation services by referring to multilingual equivalent terms. By considering machine translation services as black boxes, we achieve substantial quality improvements through Web service coordination.

Multiagent Systems are potential computational systems for various practical applications, tools, and so on. Multiagent simulation is one of the remarkable application to evaluate several kinds of phenomena. In order to design an agent for multiagent simulation, it is important to reflect user&apos;s opinion. However, if a user is not computer professional or does not have technical knowledge of agent logics and programming language, it is hard for him/her to implement his/her own opinion. Participatory design is a promising approach to incorporate user&apos;s opinion in the agent design and modification process. In this paper, we propose rehearsal oriented testing for implementation of participatory design. By the rehearsal oriented testing, it becomes possible to carry out anytime modification of agent&apos;s scenario, which describe its behavior during simulation. For rehearsal oriented testing, we set operators for modifying scenarios, which is described using finite state machine model. We also design interaction protocol between a user and an agent to smoothly get information through the user-agent dialog for modifying operators. Under this protocol, an agent informs a user about what kind of information is required.

A method to calculate the semantic dissimilarity in two countries' pictogram interpretations is proposed. Two countries' pictogram interpretation words are mapped to SUMO classes via WordNet2SUMO. Appropriate concept weights are assigned to SUMO classes using the interpretation ratios. The edges between the two SUMO classes are counted to obtain the path length of the two classes. Three bipartite graphs are generated using the classes and edges to calculate the between-country vs. within-country dissimilarity in pictogram interpretations. Preliminary result showed that human assessment of interpretation dissimilarity does not always correspond to concept-level dissimilarity in the ontology.

Navigation services for pedestrians are spreading in recent years. Our approach to provide personal navigation is to build a multi-agent system that assigns one guiding agent to each human. This paper attempts to demonstrate a design implication of the guiding agent. In the navigation experiment where a pedestrian using a map on a GPS-capable cellular phone was guided by a distant navigator, we observed the communication between them by conversation analysis. The result suggests that information required by a pedestrian were the current location, the current direction and a proper route toward a destination. The communications between a pedestrian and a navigator were based on a navigation map or a movement history. When a pedestrian did not understand the map adequately, navigation sometimes failed due to the lack of communication basis.

We propose how to acquire driver&apos;s individual operation models using the three-dimensional driving simulator in order to implement distinct personalities on each agent. In this paper, operation models are defined as sets of prioritized operation rules, each of which consists of the world as observed by a driver and his/her next operation according to the observation. Each driver might have different set of rules and their priorities. We apply a method to acquire individual operation models using hypothetical reasoning. Because of the method, we are able to obtain operation models which can explain driver&apos;s operation during driving simulation. We show some operation models acquired from aged/young human drivers, and then clarify the proposed method can catch each driver&apos;s characteristics.

To increase the accessibility and usability of online language services, this paper explains the Language Grid, a service oriented collective intelligence initiative. The Language Grid is an infrastructure that allows end-users to create new language services for their intercultural / multilingual collaboration activities. To this end, language resources including data and programs are wrapped as Web services so that users can easily share and combine these Web services to create Web service workflows that support their own collaboration tools. © 2008 IEEE.

Multilingual dictionaries are very useful in machine translations and natural language processing. However, a multilingual dictionary including all natural languages still does not exist. In this paper we propose a trustworthy method to automatically create multilingual dictionary represented by N-lingual synonymous word sets (N-tuples, hereafter). Based on the work of 3-lingual synonymous word sets, our method has extended 3-lingual to n-lingual synonymous word sets from multiple bilingual dictionaries. By matching and combining the triples instead of the binary relations in the bilingual dictionaries, the complexity of the problem is significantly reduced. Using this method, we created 4-lingual synonymous word sets among Chinese, Japanese, English and German. The evaluations indicate that our combining algorithm has effectively solved the error accumulation problem and achieved a very promising quality. In the example application, the 4-tuples are used to refine the translation quality of a multi-hop machine translator created on the Language Grid. It shows that utilizing the handy online services and uniform platform in research work is a good methodology. © 2008 IEEE.

Globalization and advanced information technologies cause the increasing development of collaborative business. This paper proposes an interorganizational workflow modeling approach for collaborative business based on local process views. In the proposed model, instead of sharing a pre-defined global workflow view among organizations, each organization has its own understanding and preference of the whole collaboration and can manage its local workflow process distributively and autonomously. Further we provide a methodology for interorganizational workflow collaboration by two phases: the bottom-up coordination phase and the top-down implementation phase. We illustrate the effectiveness of the proposed model and collaboration approach by a case of collaborative software development process.

Public Web services are not designed to be used with specific other Web services in a composite Web service. This leads to the following requirements for the proper control of a composite Web service; 1) adaptation to changes in Web services, 2) coordinating the contexts of internal processing, 3) flexible execution of human tasks. These controls must be applied following the policies of stakeholders, such as service providers. Some previous works proposed a method that adds processes to a composite Web service. Unfortunately, they fail to implement the controls needed and do not consider the policies of stakeholders. In this paper, we propose a meta-level architecture named Service Supervision, which controls a composite Web service based on the policies of stakeholders in order to achieve the requirements. To show the effectiveness of our architecture, we mention applications that currently apply Service Supervision to composite Web services consisting of machine translator Web services and human tasks.

Multilingual expression services (MESs) are available in many multicultural fields, such as education, medical care, and disaster prevention. In Japan, for example, local governments and NPOs create parallel-texts and provide expression services in such fields. However, these services present a common difficulty: the lack of parallel-texts. Service-computing technology has the potential to allow them to collaborate. In this paper, we show how Service Oriented Architecture (SOA) can be applied to multicultural activities. As a first step, we propose an architecture, in which language resources are available as web services, and MESs can obtain parallel-texts from the language web services. To achieve this, RDF metadata are attached to parallel-texts and stored in a separate RDF repository, with which users can describe what they need in detail. This arrangement allows MESs to select only the contents they need from the shared resources and show them to users. Our implementation confirms that the proposed architecture works well with two real services in use.

In this paper, we develop a novel participatory simulation platform; called gumonji/Q, by integrating scenario description language Q and network game gumonji. In a participatory simulation, humans and software-agents coexist in a shared virtual space and jointly perform simulations. To observe practical behaviors of humans, a participatory simulation platform must be able to provide reasonable simulated experience for humans to let them behave as they do in the real-world. gumonji/Q makes it possible to design diverse interaction protocols based on Q's scenario description ability. Under the "game-quality" graphics provided by gumonji, humans and agents can interact with their surrounding environment, which means they can affect the environment and receive feedback from the environment. Since gumonji/Q inherits gumonji's features as a network game, users are more enticed to participate in simulations since simulations on gumonji/Q seems more enjoyable than normal simulations. We show an example of how to obtain human behavior models through a simulation on gumonji/Q.

This paper reports on our experiences with combining Heart of Gold and Language Grid technology to provide more language resources available on Web. Heart of Gold is known as middleware architecture for integrating deep and shallow Natural Language Processing components. The Language Grid is an infrastructure built on top of the Internet to provide distributed language services. Having Heart of Gold available as Web services in the Language Grid environment would contribute to interoperability among language services. ©2008 IEEE.

Participants at both end of the communication channel must share common pictogram interpretation to communicate. However, because pictogram interpretation can be ambiguous, pictogram communication can sometimes be difficult. To assist human task of selecting pictograms more likely to be interpreted as intended, we propose a semantic relevance measure which calculates how relevant a pictogram is to a given interpretation. The proposed measure uses pictogram. interpretations and frequencies gathered from a web survey to define probability and similarity measurement of interpretation words. Moreover, the proposed measure is applied to categorized pictogram interpretations to enhance retrieval performance. Five pictogram categories are created using the five first level categories defined in the Concept Dictionary of EDR Electronic Dictionary. Retrieval performance among not-categorized interpretations, categorized and not-weighted interpretations, and categorized and weighted interpretations using semantic relevance measure were compared, and the categorized and weighted semantic relevance retrieval approach exhibited the highest F, measure and recall.

Important aspects of information technology for promoting the circulation of knowledge include human interfaces to communicate knowledge, knowledge searches, collaboration based on knowledge sharing in fieldwork, and reliable high-speed computing infrastructures. In this panel, four panelists presented the aim and goal of the research conducted under our Kyoto University Global COE program, and overviews of the activities of the four (education and research) cores: "primordial knowledge models'; "knowledge searches", 'field informatics" and "knowledge grid computing" based on interdisciplinary research.

The Language Grid is a middleware with which people can connect and use language resources such as machine translations, morphological analyzers and others created in the fields of intercultural collaboration. The Language Grid cannot exist without the collaboration of Language Grid Users who provide language and computation resources, language services, and collaboration tools. This paper overviews Language Grid Association, a user group of the Language Grid and a body promoting action research to support the multicultural society.

The Language Grid is an infrastructure for enabling users to share language resources developed by language specialists and end-user communities. The users employ the resources to support their intercultural/multilingual activities. However, there is a big gap between functions provided by the language resources and those needed by the users. To bridge the gap and satisfy users' needs, this paper extends the Language Grid by adding four language service layers: system adaptation layer, combination layer, coordination layer, and user adaptation layer. By introducing the four-layers, the Language Grid can coordinate several language resources to create new language services customized to users' activities, such as back translations, multi-hop translations, and specialized translations.

Real-time interactive virtual classroom with tele-education experience is an important type of distance learning, while the current available systems are not able to connect different classrooms in open network for intercontinental and intercultural learning. Open Smart Classroom, which is an upgrade learning system based on Smart Classroom using Web Service technology, provides more extensible and scalable features to tackle the new requirements and challenges of distance learning. Open Smart Classroom is developed based on Open Smart Platform, the software computing infrastructure for Smart Space, which provide three new features: 1) open and standard interface for better mobile devices connection and communication without any prior configuration; 2) open services invocation channel between inside modules and outside systems 3) open network in which multiple Smart Spaces can connect and communicate with each other. Making use of these new features, Open Smart Classroom shows a novel and interesting experience to both of the teachers and students for intercultural and intercontinental distant learning, which also gives a significant research perspective of future distance blended learning system.

エージェントを用いたシミュレーションには，2つの目的がある．第一は複雑な社会現象を解明するためのものである．エージェントは可能な限り単純化してモデル化され，エージェント相互のインタラクションによって複雑な現象が生まれることを観察する．第二は複雑な人間の挙動を現実に近い形で再現するものである．エージェントは実際に近い形でモデル化され，利用者に疑似体験を与える. 後者は複雑なシステムを実装する前段の実証実験や訓練に用いられる．本稿では，マルチエージェントシミュレーションを上記の両面から解説する．マルチエージェントシミュレーションを基礎技術とする社会シミュレーションは 日欧米の研究活動を統合し国際会議が生まれている．その応用は 証券取引，交通制御 航空演習，避難誘導など多岐に渡る. マルチエージェントシミュレーションはまた 参加型デザインと融合し 新しいシステム開発のツールへと発展しつつある.

エージェントの研究が世界的な広がりを見せ始めてから，今年でほぼ15年が経過しようとしています．1990年代の半ばから，マルチエージェントや自律エージェントに関する国際会議が開催されるようになりました．それらが21世紀に入り統合され，現在はAAMASという，700人ほどの参加者をかかえる大きな国際会議に発展してきています．国際会議の規模は大きくなりましたが，同時に，研究が非常に多様化・細分化し，互いの関係が複雑で理解が困難になってきていると感じています．これだけ大きく育ってきた研究分野をさらに発展させてゆくためには，こういう技術がどのような未来を切り拓くのか，また，どういう社会的インパクトがあるのかということを議論し，説得力のある説明を社会に示してゆくことが必要ではないかということで，このパネル討論を企画しました．今回のパネル討論には，エージェント研究において，日本が世界に対して重要なメッセージを発信している6つの領域を選び，各領域の代表的な研究者の方々にお集まりいただきました．

One of the most important attributes of social agents is the ability to fluently interact with human communities. Interaction scenarios are seldom designed by computer professionals, and instead are most often specified by application designers such as sales managers, travel agencies, and school teachers. Our challenge is to extend agent technologies to ensure the social acceptance of agents in various applications. To this end, we have developed Q, a language for designing interaction scenarios among agents and human communities. Q can also act as an interface between computer professionals and application designers. This paper details the specifications of Q, and shows how to design and execute scenarios for social agents.

Social law is perceived as evolving through the competition of individual social strategies held by the agents. A strategy with strong authority, accepted by many agents, will tend to diffuse to the remaining agents. The authority of a social strategy is determined by not only the number of but also the collective social positions of its overlaid agents. This paper presents a novel collective strategy diffusion model in agent social law evolution. In the model, social strategies that have strong authority are impressed on the other agents. The agents will accept (partially or in full) or reject them based on their own social strategies and social positions. The diffusion of social strategies proceeds in a series of steps and the final result depends on the interplay between the forces driving diffusion and the counteracting forces.

To test large scale socially embedded systems, this paper proposes a multiagent-based participatory design that consists of two steps; 1) participatory simulation, where scenario-guided agents and human-controlled avatars coexist in a shared virtual space and jointly perform simulations, and the extension of the participatory simulation into the 2) augmented experiment, where an experiment is performed in real space by human subjects enhanced by a large scale multiagent simulation. The augmented experiment, proposed in this paper, consist of 1) various sensors to collect the real world activities of human subjects and project them into the virtual space, 2) multiagent simulations to simulate human activities in the virtual space, and 3) communication channels to inform simulation status to human subjects in the real space. To create agent and interaction models incrementally from the participatory design process, we propose the participatory design loop that uses deductive machine learning technologies. Indoor and outdoor augmented experiments have been actually conducted in the city of Kyoto. Both experiments were intended to test new disaster evacuation systems based on mobile phones.

絵文字メッセージをやり取りして人間同士が意思疎通するには,コミュニケーションに参加している人間たちが同じ絵文字解釈(意味理解)を持たなければならない.絵文字コミュニケーションシステムの潜在利用者である日本と米国の子供たちが,同じ絵文字解釈を持つかを調べるために,システムで利用される120個の絵文字の意味を問うアンケート調査を行った.結果,ジェスチャー,色と性別,時間,空間,馴染みのもの,顔/表情の六つの領域に関係する19個の絵文字に対して,両国の子供が異なる解釈を持つことが分った.ここでは,多様な解釈を持つ絵文字を,絵文字コミュニケーションシステムでうまく利用するための方法として,絵文字の解釈語リストと頻度情報を利用し,たくさんの絵文字の中から検索語に関連する絵文字を検索する方法を提案する.提案手法は,絵文字の意味的なあいまい性を考慮した検索を実現し,文化的な解釈の差を扱う検索システムの土台となる.

The possible use of machine translation languages in understanding different languages spoken by different people in the world, is discussed. People with different cultural backgrounds learn to understand and enjoy the regional cultures they encounter, and preserve other region's indigenous cultural traditions. Machine translation systems developed for written text don't translate spoken languages well. When the error created by this system isn't acceptable, the user needs to rephrase the input text message. The scientists are trying to create communicating cultures with simple conversational agents, by moving away from metaphors such as transparent translation channels or omniscient agents. This will result in intercultural experiences to be preserved all around the world.

機械翻訳を介したコミュニケーションを通じて相互理解を実現するためには，翻訳精度の向上とともに相互作用性の向上が重要である．我々は機械翻訳に対するユーザの適応行動の1 つである原文の書き換えに注目した．本論文では，ユーザが母国語だけを用いて原文の書き換え作業を行う方法として折り返し翻訳を検討し，折り返し翻訳を用いてユーザが書き換え作業をする際の作業量を減らす支援方法を考案する．本研究でユーザの折り返し翻訳作業に関する実験を実施，分析した結果，以下の知見を得た．1) 母国語に関する知識が豊富なユーザほど機械翻訳に容易に適応でき書き換え作業量が少なかった．2) ユーザに事前に「良い翻訳結果を得るためのルール集合」の教示を行うと，母国語に関する知識が豊富でないユーザも機械翻訳に容易に適応できるようになり，書き換え作業量が大幅に減った．3) ただし，原文をどのように変更すべきかを明示しない「操作自由型ルール」に対する教示効果は薄く，これらのルール獲得にかかる書き換え作業量は大きく減少しなかった．原文をどのように変更すべきかを明示した「操作指示型ルール」に対する教示効果は高く，これらのルール獲得にかかる書き換え作業量は大きく減少した．4) ルールの教示は，母国語に関する知識が中位のユーザに最も効果的であった．Translation refinement is often observed when users communicate via machine translation systems. In this study, we analyzed user's translation refinement process through a controlled experiment. In the experiment, users translated sentences using a Japanese-English-Japanese turn-back translation. From the analysis, we discovered the following results: 1) The more knowledge users had about the source language, the better users could refine the original text, 2) Rule instruction was very effective in user's adaptation. Users who were reminded of the rules refined the original text ahead of other users, 3) Instructing operational rules were effective in helping user's adapation, while conditional rules were not as much effective. 4) Rule instruction was most effective to those who had midium knowledge in their source languages.

To increase the accessibility and usability of online language services, this paper proposes the language grid to create composite language services for various communities. The language grid is called "horizontal," when the grid connects the standard languages of nations, or "vertical," when the grid combines the language services generated by communities. Semantic Web service technologies are applied in a human-centered fashion, to create composite language services through the collaboration of users and agents. Three example scenarios are given to illustrate how the language grid will organize standard and community language services for intercultural collaboration activities.

Various situations in a massively multi-agent simulation will emerge in a simulation or the period of the simulation will become too long. These situations cause problems for system operators in that each action scenario becomes too complex to maintain and a simulation costs very long time. Therefore, flexible control of the simulation, such as changing simulation speed and switching agents' action scenarios, is required. We propose a meta-scenario description language and a meta-level control architecture. The meta-scenario description language describes how to control simulations and agents based on an extended finite state machine. Meta-level control architecture achieves control on the basis of meta-scenarios provided by a meta-scenario interpreter, which controls interpreters of agents' action scenarios and the simulation environment. In addition, our proposed architecture does not lose scalability of massively multi-agent systems for some applications.

A framework is proposed to improve Web Service performance based on context-aware communication. Two key ideas are introduced to represent a client context; (1) available protocols that the client can handle, and (2) operation usage that shows how the client uses Web Service operations. We call our context aware approach a Situated Web Service (SiWS). We implemented and evaluated the SiWS and found that the overall performance was improved if more than three Web Services were executed between context changes.

Agents will adopt different strategies in the multiagent systems. However, the strategies of agents may produce conflicts. While agents coordinate with each other in the operations, they will negotiate about their strategies to reduce conflicts. This paper models the agent strategy negotiation mechanism in three kinds of agent coordination forms: order, cooperation and non-cooperation. In the order form, an agent can order another agent to do something, so the junior agent will follow the strategy of the superior agent
in the cooperation form, two agents may cooperate to behave, so they will negotiate about their social strategies in the operations
in the non-cooperation form, the agents may behave without considering other agents' behaviors, so the agents will adopt their own respective favorite strategy. With the presented framework, the strategy negotiation among different agents can be modeled well. © 2006 IEEE.

Even though multilingual communities that use machine translation to overcome language barriers are increasing, we still lack a complete understanding of how machine translation affects communication. In this study, eight pairs from three different language communities - China, Korea, and Japan - worked on referential tasks in their shared second language (English) and in their native languages using a machine translation embedded chat system. Drawing upon prior research, we predicted differences in conversational efficiency and content, and in the shortening of referring expressions over trials. Quantitative results combined with interview data show that lexical entrainment was disrupted in machine translation-mediated communication because echoing is disrupted by asymmetries in machine translations. In addition, the process of shortening referring expressions is also disrupted because the translations do not translate the same terms consistently throughout the conversation. To support natural referring behavior in machine translation-mediated communication, we need to resolve asymmetries and inconsistencies caused by machine translations. © Copyright 2006 ACM.

The task of automatically composing Web services involves two main composition processes, vertical and horizontal composition. Vertical composition consists of defining an appropriate combination of simple processes to perform a composition task. Horizontal composition process consists of determining the most appropriate Web service, from among a set of functionally equivalent ones for each component process. Several recent research efforts have dealt with the Web service composition problem. Nevertheless, most of them tackled only the vertical composition of Web services despite the growing trend towards functionally equivalent Web services. In an attempt to facilitate and streamline the process of horizontal composition of Web services while taking the above limitation into consideration, this work includes two main contributions. The first is a generic formalization of any Web service composition problem based on a constraint optimization problem (COP); this formalization is compatible to any Web service description language. The second contribution is an incremental user-intervention-based protocol to find the optimal composite Web service according to some predefined criteria at run-time. Our goal is i) to deal with many crucial natural features of Web services such as dynamic and distributed environment, uncertain and incomplete Web service information, etc; and ii) to allow human user intervention to enhance the solving process. Three approaches are described in this work, a centralized approach, a distributed approach and a multi-agent approach to deal with realistic domains.

When multiagent simulations are used for consensus building among stakeholders, it is important not only that the domain experts can deeply understand stakeholders' actual behavior but also that the stakeholders can feel the simulation result as their solution. To this end, we propose a modeling methodology which combines several techniques with the participatory method which takes stakeholders into the modeling process using role playing games (RPG). There are two types of model required to simulate a social system as a multiagent system: agents (internal models) and interactions. Hence, we considered a modeling method according to each character. In modeling an agent (e.g. decision making) which is implicit in human, the identification of the model greatly depends on the modeler's ability. Therefore we propose a modeling method wherein classification learning creates an alternative model from RPG log data for validating the domain experts' hypothesis. On the other hand, in modeling interactions (e.g. negotiation) which are emerged outside of human, it is rather important to show and capture continuously appeared interactions. Therefore we propose a modeling method with participatory simulation where a stakeholder participates as an avatar and agents act as the other stakeholders in order to deeply understand the stakeholders' interactions. Our methodology was effective to give the domain experts a deeper understanding through a real case study of agricultural economics in the northeast of Thailand [17]. Copyright 2006 ACM.

Multilingual communities using machine translation to overcome language barriers are showing up with increasing frequency. However, when a large number of translation errors get mixed into conversations, users have difficulty completely understanding each other. In this paper, we focus on misconceptions found in high volume in actual online conversations using machine translation. We first examine the response patterns in machine translation-mediated communication and associate them with misconceptions. Analysis results indicate that response messages to include misconceptions posted via machine translation tend to be incoherent, often focusing on short phrases of the original message. Next, based on the analysis results, we propose a method that automatically predicts the occurrence of misconceptions in each dialogue. The proposed method assesses the tendency of each dialogue including misconceptions by calculating the gaps between the regular discussion thread (syntactic thread) and the discussion thread based on lexical cohesion (semantic thread). Verification results show significant positive correlation between actual misconception frequency and gaps between syntactic and semantic threads, which indicate the validity of the method. Copyright 2006 ACM.

To increase the accessibility and usability of online language services, this paper proposes the language grid to create composite language services for various communities. The language grid is called "horizontal, " when the grid connects the standard languages of nations, or "vertical, " when the grid combines the language services generated by communities. Semantic Web service technologies are applied in a human-centered fashion, to create composite language services through the collaboration of users and agents. Three example scenarios are given to illustrate how the language grid will organize standard and community language services for intercultural collaboration activities.

Among various metadata design difficulties, 1) designing subsumption relation of two classes, 2) designing diverse properties of a given class, and 3) determining whether an item should be designed as class or property, are three common difficulties faced by human designers. We devised a metadata referring method, which provides example-based design information by organizing existing metadata defined in multiple ontologies in table formats. Our referring method provides 1) existing class hierarchy definitions, 2) similar and specific property definitions, and 3) overviews of class hierarchy and property definitions of a given class, after searching through numerous application ontologies. An evaluation experiment has revealed that our method is useful in design difficulties 1) and 2), but not in 3).

Previous works on information extraction from tables make use of prior knowledge such as a cognition model of tables or lexical knowledge bases for specific domains. However we often need to interpret table structures in each table differently and to treat lexicons in various domains to more fully utilize the broad range of tables available on the Web. The method proposed in this paper uses relations represented by structures to extract an ontology from a table. Once the interpretations of table structures are given by humans, the table structures are automatically generalized to extract relations from the whole table. We define a formal representation of generalized table structure based on the adjacency of cells and iterative structures. As the result of the comparison with a method proposed in a previous work, it was shown that our method is suited to extraction of various relations which are needed for descriptions in RDF/OWL.

Some procurement negotiations are difficult to conclude smoothly using existing information processing systems because they need to first develop fuller buying/supplying specifications; through bilateral negotiations between the buyer and supplier. In this paper, we propose an e-procurement model that provides a specification improvement mechanism, whereby a buyer can improve his/her procurement specifications using the supplier's knowledge. We conducted a simulation based on a real procurement case study to establish. the effectiveness of our e-procurement model. Simultaneous procurements using multi-agent based technology can help buyers improve their own specifications. In addition, the e-procurement model helps suppliers to enhance the order-acceptance potential.

In software/content development in intercultural environments, collaboration among participants and effective administration are crucial for overall successful projects. Various tools are used to achieve the collaboration and the administration of the project: no single groupware is sufficient for both collaboration support and management. Therefore, interoperability among various tools employed in one project is essential for efficient collaboration. To establish semantic interoperability of collaboration tools, we have designed an RDF tag set that can function as vocabulary to describe information interoperably among the tools. In this paper, we propose a tag set to describe information in intercultural projects and show how the data in collaboration tools and project management tools are integrated with those semantic tags. With the tag set and the supporting method, it is expected that the members and the administrators do not have to pay much efforts to keep the collaboration status.

This paper discusses Kleinberg's HITS algorithm (hyperlink-induced topic search) that extracts the Web community by Web inherent hyperlink analysis. The problems of the algorithm are analyzed and an improvement is proposed. For this purpose, a tool (Link Viewer) that visualizes the operation process of HITS algorithm was developed. The analysis revealed the following problem of the HITS algorithm: when there exists a page in the base set which is not related to the original topic at all and has a dense link structure, it is impossible to extract the Web community (authority and hub) matched to the original topic (topic drift problem). The authors focused only on the link analysis, and proposed the following modifications: (1) a technique in the eigenvalue calculation to consider the projection on the root subspace
(2) a technique for iterative calculation by extracting only the page from the base set which has link relations to multiple pages in the root set. A technique combining (1) and (2) is also considered. As a result, the topic drift problem is avoided for any topic with a relatively small amount of computation, and the HITS algorithm is improved by using the link information. © 2004 Wiley Periodicals, Inc.

本研究では,機械翻訳システムに対する利用者適応モデルを翻訳リペア時に有効なルール集合と近似し,利用者適応について統制実験を通じて分析する.67名の被験者に日英日折り返し翻訳を行ってもらった結果以下が明らかとなった.1)原言語知識の豊富な利用者ほど容易にMTシステムへの適応が可能である.2)ルール教示はは適応に効果があり,その効果は原言語知識レベルが中位の利用者が最も高く,続いて下位,上位となる.

Semantic Web is a challenging framework to make Web information machine readable or understandable, but it seems not enough to make human's requirements for collecting and utilizing information automatically. The Agent technology becomes hopeful approach to bridge the gap between humans and machines. Agents may be autonomous and intelligent entities that may travel among agents and human. They get the requirements from human or other agents, and offer an appropriate solution through consulting among them.
The main difference between agent and ordinary software development is the issue of coordination, cooperation and learning. This issue is very important for utilizing the web information. In this paper, we attempt to give an overview and research challenges with respect to the combination of machine learning and agent technologies with Semantic Web from the perspective of interaction as well as interoperability among agents and humans.

Communityware is for more diverse and amorphous groups of people. Compared to current groupware studies, communityware focuses on an earlier stage of collaboration: group formation from a wide variety of people. We have been working on real scale experiments on communityware for more than 8 years. This paper summarizes our experiences on real scale experiments on communityware: the Mobile Assistant Project, Digital City Kyoto, and the Intercultural Collaboration Experiment.

グループダイナミクスなどの社会現象のマルチエージェントシミュレーションを行う際に重要となるのは,エージェントあるいは人間間のインタラクションを容易に表現できることである.そこで我々は,大規模な社会シミュレーションを実現するために,応用領域の専門家が複雑なインクラクションを容易に記述できるよう,シナリオ記述言語QとIPC(Interaction Pattern Card)を開発した.これらは,従来のエージェント記述言語のように,エージェントの内部メカニズムのモデル化およびその記述に焦点を当てるのではなく,外界とエージェントとのインタラクションだけに焦点を当てている.つまり,エージェントがどのようにして振る舞うのかを記述するのではなく,エージェントに何をさせるのかという依頼を記述することでエージェント群の制御を試みる.また,このような視点の変更により,従来のプログラミングプロセスとは異なり,シナリオを記述するシナリオライタとエ-ジェントシステム開発者との協力を可能にする記述プロセスが必要となる.そこで,シナリオを記述するためのプロセスとして「語彙の決定],「シナリオの記述」,「インタラクションパターンの抽出」,「実・仮想空間での実験のインタリーブ]という4ステッププロセスを確立した.最後に,これらの提案したシナリオ記述言語とシナリオ記述プロセスの有効性を評価するために,災害での避難シミュレーションに適用し検証を行った.その結果,過去に行われた実空間での統制実験の結果をシミュレーションによって追試することに成功し,提案手法の有効性を確認した.

インターネットを介したアジアワイドな協調作業を支援するため，異文化コラボレーション実験（ICE2002）を実施した．日中韓馬の5大学からの40名の教員・学生が，日中韓馬英の5カ国語の機械翻訳を組み込んだ多言語ツール，TransBBSとTransWebを介して母国語でオープンソースソフトウェア開発を行った．今後，国際プロジェクトの現場で実用に耐える異文化コラボレーション環境の開発を目指す予定である．

In this paper we discuss problems with HITS (Hyperlink-Induced Topic Search) algorithm, which capitalizes on hyperlinks to extract topic-bound communities of web pages. Despite its theoretically sound foundations, we observed HITS algorithm failed in real applications. In order to understand this problem, we developed a visualization tool LinkViewer, which graphically presents the extraction process. This tool helped reveal that a large and densely linked set of unrelated Web pages in the base set impeded the extraction. These pages were obtained when the root set was expanded into the base set. As remedies for this topic drift problem, prior studies applied textual analysis method. On the other hand, we propose two methods which utilize only the structural information of the Web: 1) The projection method, which projects eigenvectors on the root subspace, so that most elements in the root set will be relevant to the original topic, and 2) The base-set downsizing method, which filters out the pages without links to multiple pages in the root set. These methods are shown to be robust for broader types of topics and low in computation cost.

We propose an augmented Web space and its query language to support geographical querying and sequential plan creation utilizing a digital city that is a city-based information space on the Internet. The augmented Web space involves a new approach to integrate the World Wide Web (WWW) and a geographic information system (GIS). The augmented Web space consists of home pages (HP), hyperlinks, and generic links that represent geographical relations between HPs. The generic links are created dynamically using geographical evaluation functions included in a user&apos;s search query each time one is issued. A query also includes a path expression showing how to navigate the HPs, hyperlinks, and generic links. Since the path expression is an extended regular expression, we can describe an arbitrary sequence of users&apos; search actions for navigating the augmented Web space. We have applied the proposed augmented Web space to Digital City Kyoto, a city information service system that is accessed through a 3D walk-through implementation and a map-based interface. Each time a user&apos;s query is issued through the 3D and 2D interfaces. Digital City Kyoto creates an augmented Web space, and navigates the Web information space based on the path expression in the query.

The objective of this study is to explore the cultural differences that appear in terms of Web usage. We examined the homepages of 987 individuals of homogeneous social standings from Japan, the United States, Germany and South Korea. The contents of the homepages, such as academic backgrounds, research outcomes and personal information, were found to exhibit interesting differences. Subjects bi the United States are more likely to express the identity of researchers' aspect, Japanese their whole personal identities and Koreans social status. Professors from Germany tend to use prepared templates and do not offer much information on their homepages.

For realizing a robot working in human society. interaction with humans is the key issue. We have developed a robot that interacts with humans based on visual recognition. This robot has two vision systems: an omnidirectional vision system for acquiring necessary visual information and a binocular stereo vision system. The binocular vision system indicates what the robot is looking at and is not used for locomotion. Gaze control plays an important role in human-robot interaction. This paper reports how the robots gaze influences subjects' impressions of the robot. With a statistically significant number of robot observers, we employed psychological methods, the semantic differential method (SD), and factor analysis. This careful psychological analysis of robot impressions is, the first trial in robotics, Through the experiment. ive have found that the impressions mainly consist of four factors: familiarity, environment, activity, and performance. The computer skills of subjects affect their impressions of the robot.

This paper introduces a new application area for agents in the computer interface: the support of human-human interaction. We discuss an interface agent prototype that is designed to support human-human communication in virtual environments. The prototype interacts with users strategically during conversation, spending most of its time listening. The prototype mimics a party host, trying to find a safe common topic for guests whose conversation has lagged. We performed an experimental evaluation of the prototype's ability to assist in cross-cultural conversations. We designed the prototype to introduce safe or unsafe topics to conversation pairs, through a series of questions and suggestions. The agent made positive contributions to participants' experience of the conversation, influenced their perception of each other and of each others' national group, and even seemed to effect their style of behavior. We discuss the implications of our research for the design of social agents to support human-human interaction. Copyright ACM 2000.

Real-time search has two aspects, one as an efficient search method (in a single problem solving trial), and the other as an overall problem solving architecture with learning ability (through repeated trials). In both respects, the use of inadmissible (and hence inconsistent) heuristic functions bring some merits such as improved performance, but there is no theory yet that well explains when and why these algorithms benefit from them. In this paper, as a step towards fully understand and take advantage of such nonstandard heuristic functions, we discuss the properties of LRTA* and the Moving-Target Search (MTS) algorithms under heuristic functions violating admissibility or consistency. In particular, we show (1) the completeness of MTS with inconsistent or even inadmissible heuristics, and present (2) a new proof technique for the convergence of LRTA* which is applicable regardless of consistency.

This paper summarizes our recent activities to support people to communicate with each other using public computer network systems. Unlike conventional teleconferencing systems, which are mainly for business meetings, we focus on informal communication in open organizations. So far, three different systems have been developed and actually tested.
In Socia, we introduced vision agents which act on behalf of their users in a network. To enable a meeting to be scheduled at a mutually acceptable time, we proposed the scheme called non-committed scheduling.
FreeWalk supports casual meetings among more than a few people. For this purpose, we provide a 3-D virtual space called community common where participants can behave just as in real life.
In the ICMAS'96 Mobile Assistant Project, on the other hand, we conducted an experiment in an actual international conference using 100 personal digital assistants and wireless phones. Various services were provided to increase the interactions among participants of the conference.
Based on these experiences, we are cow moving towards community-ware to support people to form a community based on computer network technologies.*

Linear-space search algorithms such as IDA* (Iterative Deepening A*) cache only those nodes on the current search path, but may revisit the same node again and again. This causes IDA* to take an impractically long time to find a solution. In this paper, we propose a simple and effective algorithm called Stochastic Node Caching (SNC) for reducing the number of revisits. SNC caches a node with the best estimate, which is currently known of the minimum estimated cost from the node to the goal node. Unlike previous related research such as MREC, SNC caches nodes selectively, based on a fixed probability. We demonstrate that SNC can effectively reduce the number of revisits compared to MREC, especially when the state-space forms a lattice.

In order to understand cognitive aspects of autonomous robots, it is interesting to develop a mechanism by which the robot autonomously analyzes physical sensor data and construct a state space. This paper proposes a coherent approach to construct such a robot oriented state space by statistically analyzing sensor patterns and rewards given as results of task executions. In the state space construction, the robot creates sensor pattern classifiers called Empirically Obtained Perceivers (EOPs) the combinations of which represent internal states of the robot. A novel feature of the method is that the EOP controls attention to select necessary information and the state space is obtained with the attention control mechanism. We have confirmed that the robot can effectively construct state spaces through its vision sensor and execute navigation tasks with the obtained state spaces in a simple simulation world.

In this paper, we discuss two major tradeoffs, spatial and temporal tradeoffs, that appear when applying market-based computing to multimedia network applications. The former appears between computation and communication cost, depending on how agents are distributed over network. The latter appears between reactiveness and correctness of result, depending on how the network environment dynamically changes.
By implementing a market-based resource allocation mechanism to a desktop conferencing system, we clarified that (1) as for spatial tradeoff the centralized computation becomes profitable in proportion to the number of clients, and that (2) as for temporal tradeoff the merit to respond quickly to the change of the environment by prematurely terminating the computation supersedes the merit to improve the accuracy of the resource allocation by performing the calculation until the market perfectly clears. It has also been proved that the market-based mechanism can achieve efficient allocation in an actual network environment.

移動通信の普及と携帯情報端末の高性能小型化により, モバイルコンピューティングが身近なものになってきた. 本稿では, モバイルコンピューテイング環境の制約について考察すると共に, モバイルアプリケーションの開発において要求される機能や課題の整理を行い, 課題解決に向けた技術としてエージェント通信技術の動向を紹介する. また, 筆者らが実施した,100台の携帯情報端末を用いた国際会議支援実験を紹介し, 更に大規模な, 数万台の端末群を想定したモバイルコンピューティング環境の実現に向けて, 技術課題を考察する.

携帯型計算機の普及・低価格化および携帯電話, PHSに代表されるワイヤレス通信手段の高性能化に伴い, モーバイル通信に対するが期待が高まっている. 今回我々はICMAS96(The Second International Conference on MultiAgent Systems)という比較的同種の目的をもって人が集まる場において, モーバイル通信サービスとして国際会議支援を行なうシステムを構築し, 評価実験を行なった. 本報告ではシステム概要について報告するとともにユーザの使用状況からモーバイル環境におけるシステムのあり方について考察する.

The state space of a sensor-based robot in the most previous works has been determined based on human intuitions, however the state space constructed from human viewpoints is not always appropriate for the robot. The robot has a different body, sensors, and tasks, therefore, we consider the robot should have an original internal state space determined based on actions, sensors, and tasks. This paper proposes an approach to construct such a robot oriented state space by statistically analyzing the actions, sensor patterns, and rewards given as results of task executions. In the state space construction, the robot creates sensor pattern classifiers called Empirically Obtained Perceivers (EOPs) the combinations of which represents internal states of the robot. We have confirmed that the robot can construct original state spaces through its vision sensor and achieve navigation tasks with the obtained state spaces in a complicated simulated world.

The capability of learning is one of the salient features of realtime search algorithms such as LRTA*. The major impediment is, however, the instability of the solution quality during convergence: (1) they try to find all optimal solutions even after obtaining fairly good solutions, and (2) they tend to move towards unexplored areas thus failing to balance exploration and exploitation. We propose and analyze two new realtime search algorithms to stabilize the convergence process. epsilon-search (weighted realtime search) allows suboptimal solutions with epsilon error to reduce the total amount of learning performed. delta-search (realtime search with upper bounds) utilizes the upper bounds of estimated costs, which become available after the problem is solved once. Guided by the upper bounds, delta-search can better control the tradeoff between exploration and exploitation.

Working from current omputer-aided instruction (CAI) systems, the authors developed the MASTERS intelligent CAI system with a general-purpose CAI control method using a production system. Two objectives were sought: (a) enabling scene display controls flexible enough to be matched to the level of understanding and characteristics of the individual student; and (b) mounting all the teaching material on a database and simplifying reuse of the scenes. In order to achieve these objectives in MASTERS, the required teaching knowledge was established as production rules and used to control scene display. Various solutions discovered to a number of problems occuring with this approach are discussed.

国際出願 PCT/JP2004/009973平成17年（2005）3月31日.