From Corpus Queries to Chat Prompts: AI in CBRP

Teachers who examine their own classroom language through Corpus-Based Reflective Practice (CBRP) stand to gain considerably, and the reflective and developmental value of the approach is well documented. Adoption remains limited, however: a single lesson can take hours to transcribe, and the corpus tools on which the analysis depends assume technical training that most practitioners never receive. This study investigates the extent to which recent artificial intelligence (AI) technologies can alleviate these barriers and facilitate the integration of CBRP into routine teaching practice. Two online English as a Foreign Language (EFL) classroom sessions involving Iranian learners were transcribed manually and using an AI-based automatic speech recognition (ASR) system (Rev.ai); the resulting transcripts were compared to evaluate the suitability of AI-generated transcription for CBRP. Subsequently, corpus analyses produced by AntConc were compared with analyses generated by three large language models (LLMs): ChatGPT Free, Claude Fable ۵, and o۳ Pro. The transcripts exposed systematic weaknesses in the ASR output: whole conversational turns were lost, code-switching into Farsi was rarely transcribed at all, Persian names appeared in corrupted form, and the interactional structure of the lessons was not preserved. Transcripts of this quality cannot support detailed corpus-based reflection. The comparison of corpus analyses showed that, although the LLMs could not reliably reproduce the frequency counts generated by AntConc, they proved strong at interpretation: they distinguished teacher from learner language, identified discourse patterns and language errors, and drew pedagogically useful insights from the data. These findings suggest that current AI technologies cannot yet replace either human transcription or dedicated corpus software, but they can substantially lower the expertise barrier to corpus interpretation and serve as powerful complementary tools within CBRP.

کلیدواژه ها:

Artificial Intelligence ، Automatic Speech Recognition ، Corpus-Based Reflective Practice ، Corpus Linguistics ، English as a Foreign Language