日本語－手話辞典の自然言語処理による分析

Abstract

Japanese–Japanese Sign Language Dictionary, issued by Japanese Federation of the Deaf, is assumed as one of the best-organized corpus for sign language researchers as well as sign language learners. The authors also utilized the dictionary to obtain new coordinate system for better sign animations via natural language processing technique as reported in previous paper [2]. On the course of the analysis, the authors are forced to digitize the dictionary to obtain notations of each sign words (denoted as sign illustration in the dictionary) written in Japanese. The very basic language operation applied to restoring recognition errors of optical character reader revealed unexpected features of the dictionary. 630 illustrations out of 3235 illustrations used in the dictionary are expressed in two or more different sentences. Although the unevenness is mostly caused by simple variation of Japanese expressions, some of the variations are caused by inflections of sign expressions themselves depend on contexts. As the unevenness may be caused by nature of sign language itself, researchers to tackle cognition mechanisms of sign language via sign notations need to take the feature account.