First International Conference on Spoken Language Processing (ICSLP 90)

Kobe, Japan
November 18-22, 1990

The Linguistic Processing Module for Japanese Text-To-Speech System

Tohru Shimizu, Norio Higuchi, Hisashi Kawai, Seiichi Yamamoto

KDD R&D Laboratories, Kamifukuoka-shi, Saitama, Japan

A linguistic processing module for Japanese text-to-speech system for use on personal computers has been developed. The linguistic processing module analyzes Japanese sentences and outputs information for the rule-based speech synthesizer. This module is implemented as software for 16-bit personal computers, running under the operating system MS-DOS. The linguistic processing module has the following two characteristics. (1) Both a morphological analysis module and a Kana-Kanji conversion module are available to analyze input texts, and the editing module can modify the output of these two modules. (2) The syntactic analysis which is introduced in morphological analysis decreases the number of word candidates, and consequently the system has high performance for selection of correct words. The result of the evaluation tests shows that the accuracy of the morphological analysis is 99.0% for the sample articles of a popular economical newspaper.

Full Paper

Bibliographic reference.  Shimizu, Tohru / Higuchi, Norio / Kawai, Hisashi / Yamamoto, Seiichi (1990): "The linguistic processing module for Japanese text-to-speech system", In ICSLP-1990, 321-324.