2023 Theses Master's

Interfaces for Personalized Language Learning with Generative Language Models

Kwon, Taeahn

People learn a foreign language to use in diverse situations. However, current language learning technology prescribes largely fixed content to all students, which often is not relevant or engaging. To enable highly personalized language learning, we propose to leverage the contextualized language knowledge encoded in large language models (LLMs). We explore the design space of LLM-enabled language learning by developing two interfaces—GPTChat and GPTutor—that uses GPT-3 to generate language examples, such as words and sentences, in response to contexts given by the students themselves. The design of each system is informed by in-depth interviews with language learners, as well as theories in language learning. We conduct preliminary evaluations of each interface to demonstrate the potential of LLM-driven systems to offer students more personalized and relevant learning material.