Symposium: Frontiers in Natural Language Processingand AI Agents

AI agents are increasingly deployed in domains ranging from personal assistants to large-scale autonomous systems. Their performance largely depends on the cognitive architectures that support perception, reasoning, planning, and action in complex environments.

In this talk, I will present a comparative overview of existing cognitive architectures for AI agents, outlining their core features, strengths, and limitations. I will then discuss how LangGraph, an open-source framework, can be leveraged to implement and evaluate diverse cognitive architectures in practice.

Through demonstrations, I will showcase how LangGraph facilitates the development of AI agents and enables systematic performance comparisons across different cognitive models.

Language education is the process and practice of teaching or learning a second or foreign language. Natural Language Processing (NLP) technologies have been widely applied to help with language teaching and assessment.

In this talk, I will focus on three educational NLP tasks: generating AI teacher responses in educational dialogues, automated assessment, and grammatical error correction.

Towards the end, I'll showcase some potential applications of Large Language Models (LLMs) for language education, and discuss the risks and ethical considerations surrounding generative AI in education technology for language learners.