Corpora first: Evidence-driven language teaching in the age of AI

December 5, 11:00-13:00 Academic Building 8^th floor, Macao Polytechnic University

Pascual Pérez-Paredes

Universidad de Murcia/The Education University of Hong Kong

This talk argues that language corpora are not yesterday’s tech but the backbone of responsible language learning in the age of generative AI. Corpora provide authentic and diverse input that trains learners’ noticing of lexis, phraseology, and pragmatics, skills that Large Language Models (LLMs) often smooth over. Corpus methods make feedback and assessment more transparent: frequency, dispersion, and register profiling let teachers and learners calibrate targets and justify revisions beyond intuition. Corpora power hybrid human-AI workflows: corpus-to-AI and AI-to-corpus. Learner-facing concordancing develops metacognitive control and critical AI literacy where students learn to question unfounded suggestions, compare variation across genres, and use evidence to support critical language learning. The talk is inspired by my most recent book Data-driven Learning in and out of the Language Classroom, published by Cambridge University Press and co-authored with Prof Alex Boulton.