During this research period, we investigated computational approaches to enhance English for Academic Purposes (EAP) writing support through three interconnected research directions: academic writing assistance, language identification, and contextual understanding.Our primary focus was developing academic writing assistance technologies. We created methods to enhance Large Language Models for sentence-level revision tasks, with results under review in the Asian-Pacific Journal of Second and Foreign Language Education. Building on this foundation, we developed AcademiCraft, a comprehensive Multi-Agent System for EAP writing assistance, currently under review in Information journal.In our second research direction, we addressed language pattern recognition challenges. We established a two-stage recognition method for Native Language Identification in ultra-short academic texts, published in the International Journal of Advanced Computer Science and Applications (2024). This breakthrough significantly improved identification capabilities for brief academic writing samples.Our third research direction explored contextual understanding applications. We employed multi-task learning approaches for bridging resolution focused on robot instructions, with findings under review in Robotica journal. This work extended to emotion recognition with our Enhanced Real-Time Emotion Detection Framework for edge devices in emotional robotics, accepted at the 8th International Conference on Artificial Intelligence and Big Data (ICAIBD 2025). Further advancing this direction, our research on Adaptive Thresholding Triplet Loss for person identification was accepted at the 9th International Conference on Intelligent Systems, Metaheuristics & Swarm Intelligence (ISMSI 2025).The results from these three research directions demonstrate practical applications of computational methods in academic writing support and educational technology, providing effective solutions to longstanding challenges in these fields.