ИСТИНА |
Войти в систему Регистрация |
|
ИПМех РАН |
||
This paper aims to demonstrate how the orthographic transcription method used in Russian Text-to-Speech systems may help to phonetize English words and word combinations occurring in Russian texts. These are mostly names of companies, mass media, different brands and computer programs. As English words are written in the Latin alphabet they must be rendered into a graphic or phonetic system that the Russian synthesizer can interpret. In our work we propose to use the English-Russian orthographic transcription in order to transform English words into the Cyrillic script. Alternatively to transliteration orthographic transcription has the advantage that it relies on word pronunciation. It also allows to phonetize English words using only the Russian sound set without extending the TTS-system with English phonemes or extra acoustic databases. After English words are presented as a sequence of Cyrillic symbols they can be treated on all following synthesis stages in the same way as normalized Cyrillic words are. So the method proposed is universal and can potentially be used by any Russian TTS-system. In this article we also describe a perceptual experiment that was carried out in order to assess the quality of English words synthesized in this way. The proposed transcription method allowed to achieve a 87% intelligibility rate of English words phonetized by a Russian TTS-program and to highlight several arguable points in the English-Russian orthographic transcription rules.