Semantic Word Dictionaries for Japanese-to-English Machine Translation

Akio YOKOO*1 Masahiro MIYAZAKI*2 Satoru IKEHARA*3 Satoshi SHIRAI*1 Satsuki ABE*4

*1 NTT Communication Science Labolatories
1-1 Hikari-no-oka, Yokosuka 239, JAPAN
{ayokoo,shirai}@cslab. kecl.ntt.co.jp
*2 Faculty of Engineering, Niigata University
8050 Ikarashi 2-nocho, Niigata 950-21, JAPAN
miyazaki@info.eng.niigata-n.ac.jp
*3 Faculty of Engineering, Tottori University
4-101 Koyamacho-minami, Tottori 680, JAPAN
ikehara@ike.tottori-u.ac.jp
*4 NTT Advanced Technology Corporation
90-6 Kawakamicho Totsuka-ku, Yokohama 244, JAPAN
satsuki@totsuka.ntt-at.co.jP

Abstract

In order to realize high-quality semantic analysis in Japanese-to-English select and collect words to build the semantic word dictionaries. We collected a total of 400,000 words, including 200,000 proper nouns and 50,000 technical terms, based on these criteria. We gave each word morphological information, semantic information such as part-of-speech, and semantic categories from about 3,000 attributes. These dictionaries are used in the semantic analysis process in the machine translation system, ALT-J/E.



key words

machine translation, Japanese-to-English translation, semantic analysis, semantic word dictionary, vocabulary



[ Technical Report of IEICE, NLC97-13, pp.37-44 (July, 1997). ]