| Summary: | 碩士 === 淡江大學 === 西洋語文研究所 === 90 === This thesis is a corpus-based investigation of a prevalent type of lexical error in written English of Taiwan learners─lexical collocation errors. The miscollocation data used for this study has revealed that one type of lexical collocation error, Verb Noun (V N) miscollocations, occupy by far the largest portion of the learners’ collocation errors. The study therefore focuses on incorrect VN collocations that learners produce; for example, the attested learner example *‘write homework’ is a typical V N collocation error. The main goals of the study are (1) to see if any patterns or generalizations can be found by looking at a large number of learners’ VN miscollocations in terms of lexical semantic relations, and (2) to gain insights into how learners develop collocational competence as well as acquire collocational knowledge.
To investigate Verb Noun collocation errors that learners produce, I explore learner essay data from a learner corpus (English TLC), and I make use of teachers’ online corrections on those essays, corrections which are automatically indexed and searchable in an online writing environment. Using the learner corpus as my data source, I uncover 233 VN miscollocations to analyze. I incorporate lexical semantic relations and verb lexicographer files of the WordNet lexical database in the analysis of the data. The results suggest that learners’ VN miscollocations are not arbitrary; instead, patterns of learners’ miscollocations exist. The probability of learners misusing a verb collocate over a noun collocate is 93% (217 out of 231 cases), and the percentage of verb miscollocates where the wrong verb happens to be either a synonym, hypernym or a hyponym of the correct verb collocate is 56% (131 out of 233). The data also suggests that L1 plays a significant role in terms of the lexical choices that learners make.
Two applications of the results are reported. One is ‘Lexical Assistant’ (still being developed). This software is designed to find the correct verb collocate for a VN miscollocation automatically. The other application is designed to implement my analysis of VN miscollocations in a grammar checker. The checker can automatically detect a subset of VN miscollocations uncovered in my data analysis at a high rate of precision. Relevant issues and future work are also reported.
|
|---|
