CorpusChat: integrating corpus linguistics and generative AI for academic writing development

Abstract

This article reports on the development and piloting of an innovative AI-empowered platform—CorpusChat—offering customised corpus-informed chatbots for university students to develop discipline-specific academic writing skills. Corpus­Chat seamlessly merges the affordances of corpus linguistics for disciplinary writing with generative artificial intelligence (GenAI), while working to address their respective shortcomings i.e. technical and user-interface issues with corpus approaches; accuracy, reliability, and hallucinations with GenAI. This synergy enables learners’ deeper understanding of authentic language corpus data while streamlining the corpus query and analysis process. We present the findings of a pilot study involving CorpusChat with an undergraduate class for arts students at a university in Hong Kong. Two chatbots were created using the British Academic Written English corpus (BAWE) and an in-house Corpus of Humanities and Arts Texts (CHAT) totalling 656,484 words. Based on data from a student survey and focus group interviews, the results suggest that combining GenAI and corpus linguistics can positively motivate students’ disciplinary writing practices and digital disciplinary literacies. Practical observations about how to harness CorpusChat for disciplinary academic English in future research are also provided.