Presentation From Data to Dialogue: Using Corpus and AI to Enhance Self-Directed English Speaking more

Technology is advancing rapidly and is increasingly encouraged for use in language teaching and learning. This study aims to design a Corpus- and AI-aided self-directed English speaking training programme, examine its effectiveness, and explore learners’ attitudes toward this approach. Thirty-one undergraduate EFL learners participated. They completed a pre-test, a five-session Corpus- and AI-supported speaking training, a post-test, a learning portfolio, a survey, and a follow-up interview. Findings showed that participants’ overall speaking performance improved after the training, along with gains in four subskills: lexical resource (LR), grammatical range and accuracy (GA), fluency and coherence (FC), and pronunciation (PN). Among these, the greatest progress occurred in LR, while PN showed the least improvement. Learners agreed that the inclusion of a spoken corpus enabled them to identify linguistic features influencing oral proficiency, and that AI tools were effective in supporting speaking practice. The AI-enhanced environment also promoted interactive, self-directed learning. Most participants expressed positive attitudes toward the combined Corpus and AI-aided approach and demonstrated willingness to continue interacting with AI tools despite some limitations. However, a few learners reported their insufficient AI literacy and concerns about AI feedback accuracy. Overall, the study provides theoretical and pedagogical insights for enhancing English speaking instruction.