イディオム検索のためのアルゴリズム開発に関する実験的研究

概要

論文の詳細を見る
The purpose of the present study was to develop a set of computer programs for retrieving English idioms from a database and to evaluate their functions. A total of nine programs which vary in the methods of retrieving idioms were experimentally developed. A sample set of forty idioms and a corpus of 392,170 sentences collected from English magazines and TV news broadcasts were created in order to examine the performance of the programs. The program which demonstrated the best performance in idiom retrieval was the one in which the constituent words fell within a seven-word sequence in a sentence. Within this seven-word sequence, the program accepted all changes in word order and grammatical forms of the constituent words. The results also indicated that by the use of the program 99.6% of time and labor can be saved in retrieving idioms from the original corpus and the accuracy of human observation can be improved by 30% in the subsequent reconfirming process.
外国語教育メディア学会の論文