Clustering, visualising and elaborating DDL-integrated error correction process in EFL writing activities

The use of language corpora for EFL (English as a foreign language) learning purposes, known as data-driven learning (DDL), has shown effectiveness from many recent studies. Although DDL seems promising, it is barely popularised in Chinese EFL context for distinct reasons, such as deficiency of process investigation and exclusivity of other available consultation resources in DDL research.

This study aims to unpack DDL-integrated EFL writing error correction in real world, where common consultation resources can be referred to along with DDL. Fifty-nine participants in a Chinese university completed six writing tasks with follow-up revisions, eleven of which then joined stimulated recalls. TraMineR, a sequence data analysis toolkit, was used to visualise consultation processes for error correction and cluster representative trajection by error types. Retrospective device provided insightful details to further explain sequential data in error correction processes.

Results showed DDL played a significant role in error correction activities, though it functioned variously regarding error types. DDL helped participants either retrieve prior knowledge or explore new linguistic knowledge with multiple cognitive strategies. Drawbacks of DDL also raised by participants, indicating the importance of combining other consultation resources in error correction activities for better performance.