RAPID DEVELOPMENT OF TEXT TO SPEECH SYSTEM FOR UNSUPPORTED LANGUAGES USING FAKING INPUT APPROACH: EXPERIMENT WITH MUONG

Abstract

Development of a text-to-speech (TTS) system for a language normally requires a big workload, it need much research on spoken language processing and also linguistic research. For minority languages or under-resourced languages (so called unsupported languages), building a TTS system is very hard work, in some cases it is impossible, due to the lack of available resources and linguistic knowledge. This paper presents the experiment of building a TTS system for Muong language, a minority language of Vietnam, following the “faking input” approach. Based on the phonetic correlation between Muong language and Vietnamese language - a closely related language, a set of phonetic transformation rules was proposed to transform the text input of Muong language to the suitable input of Vietnamese Text-To-Speech system. That allows synthesizing the Muong speech by available Vietnamese TTS systems. The Muong synthesized speech was thus evaluated by Muong peoples in perception tests. The result shows that the synthesized speech is well understood, this suggests the possibility of this approach for rapidly and simply building TTS systems for unsupported languages.