2004 Fiscal Year Final Research Report Summary
A Descriptive Study of English Collocations and a Methodological Study of Measuring the Association Strength of Collocations : A Corpus-Based Approach
| Project/Area Number |
14510514
|
|---|
| Research Category |
Grant-in-Aid for Scientific Research (C)
|
| Allocation Type | Single-year Grants |
|---|
| Section | 一般 |
|---|
| Research Field |
英語・英米文学
|
|---|
| Research Institution | Nagoya University |
|---|
Principal Investigator |
TAKEZAWA Naohiro Nagoya University, Graduate School of International Development, Associate Professor, 大学院・国際開発研究科, 助教授 (60252285)
|
|---|
| Project Period (FY) |
2002 – 2004
|
| Keywords | collocations / corpora / MI-score / t-score / co-occurrence of words / co-occurrence of words and constructions / collocations of -ly adverbs and adjectives / corpus tools |
|---|
| Research Abstract |
Owing to the development of computer technology, large-scale corpora have been made available since the mid-1990s. Use of large-scale corpora has revealed a lot of new linguistic facts. One of the linguistic studies which corpora have made possible is research on collocations using proper tools for measuring the association strength between words. For example, it has been noted in the pre-corpora study that collocations like "abundantly clear" and "a whole new way" are natural while "abundantly hot" and "a whole cold day" are impossible, but a corpus-based study has made it possible to objectively measure the association strength between words. The MI-score (the Mutual Information score) and the t-score have been proposed. In this study, we have conducted detailed descriptions of collocations using large-scale corpora of English and discussed how to use these statistical measures for retrieving significant collocations. The corpora used for research are : The British National Corpus and The Bank of English corpus.
We have paid special attention to collocations of -ly adverbs and adjectives, and have retrieved significant collocations using the MI-and t-scores as well as the frequency. We have also discussed how to use these measures.
Our primary focus is on the co-occurrences of words, but our attention has also turned to the co-occurrences of words and constructions. The SOV construction and the "haven't NP" pattern have been analyzed from this viewpoint, and lexical patterns which frequently appear in them have been identified.
|
