Coding and theme patterns.

Despite the growing interests in investigating the application of data-driven learning (DDL), much existing research remains outcome-oriented. Limited attention has been paid to learners’ interactions with corpora, especially the experiences of consulting corpora and decision-making processes during revision in second language (L2) writing. In this regard, this study investigates how corpora assist language learning during the revision process in a classroom-based foreign language learning context. We recruited 123 non-English major undergraduates from a university in China, learners’ revision behaviours and outcomes were analysed in each draft. To gather the learners’ experiences and perceptions of using the BAWE corpus for error correction, a stimulated recall interview was conducted using the revised drafts as stimuli. Quantitative results indicated that corpus consultation was highly effective for revising word form errors (93.6%), and participants were more likely to use the BAWE corpus to correct collocation and phrase errors (62.7%). Interview results demonstrated that the corpus-based error correction method was effective in enhancing correctness and facilitating language learning. Learners were able to utilise corpora to identify patterns, shape writing habits, and test linguistic hypothesis. The findings and the implications of this study also provide valuable insights for teachers into the potential role and implementation of corpora in designing classroom writing tasks.