Integrating AI technology into corpus-based language learning through ChatAI

Abstract

This paper proposes a novel solution to the problem of integrating AI technology with traditional corpus methods through the introduction of a ChatAI tool within the AntConc corpus analysis toolkit. ChatAI is designed to simplify both the management and use of large language models (LLMs) in classroom settings. On the management side, the tool allows teachers and learners to easily access and use powerful propriety LLMs, such as OpenAI’s GPT5 models. It also allows them to download and run open access and open weight LLMs, such as Llama models, on their own computers, while keeping their data completely private. On the language learning side, ChatAI offers users a way to directly interact with LLMs both when creating corpus queries and also when analyzing or interpreting the results generated from corpus queries. As a result, teachers and learners can begin a lesson by querying a corpus with natural language using a full range of traditional corpus methods (e.g. Key-Word-In-Context, Dispersion Plots, Clusters, N-Grams, Collocates, Word/Keyword, and Wordclouds), and then pipe the results of their analyses to an LLM for further insights and discussion. Importantly, in contrast to online services like ChatGPT, ChatAI improves transparency by allowing users to see and edit important LLM settings, such as the system and user prompts as well as the temperature setting. These features not only lead to a smooth and intuitive lesson flow, but also improve the authenticity and replicability of results, while at the same time greatly reducing the possibility of LLM hallucinations.