Data-driven learning of collocations by Chinese learners of English: a longitudinal perspective

Abstract

Collocations, a crucial component of language competence, remain a challenge for L2 learners across all proficiency levels. While the data-driven learning (DDL) approach has shown great potential for collocation learning from a shorter-term perspective, this study investigates its effectiveness in the long term, examining both linguistic gains and changes in learners’ confidence about which words collocate. The effect of DDL was compared with learning collocations using a corpus-based collocations dictionary and non-corpus-based tools. Learners’ experience with the tools was also explored through a questionnaire. The study employed a quasi-experimental research design with 100 Chinese learners of English as participants in two experimental groups and a control group. A novel corpus tool, #LancsBox (Brezina et al., 2015), was used in the DDL approach to identify and visualise collocations. The results showed that the learners in the DDL group increased their collocation knowledge at the end of the treatment and retained the gains three months later. The learners also reported a significant increase in their confidence about which words collocate. Both changes were found to be more substantial than the effects of using the corpus-based collocations dictionary or other tools. As for their experience, learners reported satisfaction with using corpora in their writing and, importantly, continued with corpus consultation three months after the end of the intervention. The findings have implications for integrating corpus consultation into learning practice both inside and outside of the classroom, showing that with sufficient training, DDL can provide an effective method to learn complex linguistic features such as collocations.