英語コーパス解析に基づく類似名詞と共通共起動詞の検出 : コロケーション辞典への活用

抄録

The aim of this paper is to develop a method for classifying English nouns according to the commonality of co-occurrence verbs from a lexicographical perspective. A one-million-word corpus created at the Nagoya Institute of Technology was parsed, one sentence at a time, using a syntactic parser called Machinese Syntax. The occurrences of Noun- Verb collocations obtained from the parsed output were counted under each syntactic category. Categories were a subject-verb construction and a verb-object construction, among others. Based on the distributional hypothesis, the similarities between each noun were calculated according to the similarity between the patterns offrequency distribution of co-occurrence verbs. For many nouns, appropriate sets of semantically similar nouns that shared many co-occurrence verbs were found. As an example, it was shown that method, technique, and approach share use, apply, base, obtain, develop, and so on. No existing collocation dictionaries have provided such information on the commonality of co-occurrence verbs between semantically related nouns to date. This method can be used effectively to build a useful collocation dictionary.