日英双方向機械翻訳システム BIJET の日本語処理と翻訳実験について

概要

論文の詳細を見る
The authors have recently constructed an experimental machine translation system named BIJET (<BI>____directional J___apanese E___nglish T___ranslation system). This system consists of three parts : a Japanese processor, a English processor and a transfer processor. Both Japanese and English processors are divided into a morphological analyser, a morphological synthesizer and a syntactic processor. A transfer processor is divided into a Japanese-English transfer processor and a English-Japanese one. This system has the following features, (1) BIJET works in two modes : translation from Japanese to English and from English to Japanese. The Japanese syntactic processor works as analyser in Japanese-English translation mode and as a synthesizer in English-Japanese mode. The English syntactic processor works in the same way. (2) This system uses transfer method. Its intermediate representations are English D-tree and Japanese D-tree. Japanese D-tree and linguistic information in Japanese are newly proposed. Japanese D-tree is constructed from dependency relation and RENSETSU rule. In transfer processors, conversions,"transfer", are made between Japanese and English D-free. This paper gives the details of the Japanese processor and the transfer processor in BIJET, then discusses translation experiments and its results.
九州大学大学院総合理工学研究科の論文
1985-01-01