Exploring the effects and use of a Chinese–English parallel concordancer

Abstract

Previous studies on self-correction using corpora involve monolingual concordances and intervention from instructors such as marking of errors, the use of modified concordances, and other simplifications of the task. Can L2 learners independently refine their previous outputs by simply using a parallel concordancer without any hints about their errors and guidelines from instructors? This paper addresses this question and reports a pilot study about the effects and use of a parallel concordancer in a Chinese–English translation task. Twenty-one subjects, whose English reading proficiency attained higher-intermediate level, participated in the study, in which they used the Yahoo Chinese–English bilingual dictionary in the pretest and employed the CERT parallel Chinese–English concordancer to modify their translations in the posttest. The results indicated that the mean score in the posttest was significantly higher than that in the pretest, suggesting that learners can independently make corrections by consulting parallel corpora without any simplification of the task as well as hints and guidelines from their instructors. Inspection of the subjects' translations showed that the improvements made in the posttest centered on phraseology involving word choices and word combinations. The users' attitudes toward CERT as well as the theoretical implications of parallel concordancers in second language acquisition theories are discussed.

Keywords:

• parallel concordancer
• bilingual concordancer
• data-driven language learning
• bilingual phraseology
• learner autonomy
• parallel corpora
• query expansion
• concept-based retrieval
• translation corpus; comprehensible input hypothesis; comprehensible output hypothesis

Acknowledgements

The author would like to thank his research assistants Shen Ding, Huang Yu-Ming, and Huang Tz-Huan for helping to implement the CERT parallel concordancer. Special thanks go to the two anonymous reviewers, Jason Chang, Liou Hsien-Chin, Chen Hao-Jan, Liao Meei-Ling, Liu Chao-Lin, Ted Partington, Emma Brooks, and the Consulting Center for Statistics and Bioinformatics at National Taiwan University. The research reported in this paper is supported by research grants from the National Science Council of Taiwan under the contracts of NSC 95-2411-H-002-045-MY2 and NSC 98-2410-H-002-121.