2013 Fiscal Year Final Research Report

Study on Adaptation of Speech Synthesis for Minority Languages

Research Project

Project/Area Number	23500170
Research Category	Grant-in-Aid for Scientific Research (C)
Allocation Type	Multi-year Fund
Section	一般
Research Field	Intelligent informatics
Research Institution	The University of Tokyo
Principal Investigator	IWABUCHI Mamoru 東京大学, 先端科学技術研究センター, 准教授 (80335710)
Project Period (FY)	2011 – 2013
Keywords	自然言語処理 / 音声合成
Research Abstract	This study pursued an economical method to create a text-to-speech (TTS) software for a new language using an existing voice synthesizer. The new sound conversion rules were created with manually edited letter-to-sound rules by native speakers for most frequently used words. In order to evaluate the method, Japanese and Nepali TTS systems were created using Hindi TTS that was commercially available. The created Nepali TTS produced well recognized speech with 60-80 percent accuracy. Meanwhile the created Japanese TTS produced speech that was recognized with less than 30 percent accuracy. It was concluded that the method of this study could be usable for a language pair where the original and target minority language are phonetically and linguistically close. However there would still be challenges about language-specific issues, such as prosodic modeling, unknown words pronunciation, and grammatical analysis.