81006696

open access

Version of Record

Uncovering Collocation Errors by Using Automatic Collocation Extraction and Comparison

Learner Corpus Studies in Asia and the World

2

137-147

神戸大学国際コミュニケーションセンター

2014-05-31

2014-06-25

Collocations have been widely recognized as an essential component of the lexical competence. Many studies, however, have consistently revealed that 1.2 learners had insufficient collocational knowledge. Given its importance and learning difficulty, many researchers are interested in gaining more thorough information about collocation errors. Most previous studies, however, are based on manual analysis on small learner corpora. For researchers who work on a larger learner corpus, the manual analysis method. would be impractical. To overcome the aforementioned limitations, an innovative extraction method is proposed. This automatic retrieval approach is achieved through the use of The Sketch Engine, a corpus system developed by Adam Kilgarriff and his associates. In this study, a 1.3-miliion-word Taiwanese learner English corpus was first uploaded onto the SKE for automated comparison with a 90-million-word native English corpus (The written corpus of BNC). Based on this machine-aided collocation comparison method, many types of high-frequency V-N collocation errors were identified. The Sketch Engine tools were found to be very robust in uncovering various miscollocations in a learner corpus.

紀要論文