A keyvowel approach to the synthesis of regional accents of English.

摘要

Most English text-to-speech synthesisers offer one ofudonly two accents: General American or RP. Developinguda new accent is laborious, since it is not possible toudchoose one accent as a base form and systematicallyudtranslate to others. We use the approach of Wells ([1]),udcategorising vowels in terms of abstract keywords thatudencode classes of words. Thus it is unnecessary to use audphonemic transcription in either the development or theudexecution of a synthesiser. The “keyvowel” system canudbe used throughout the synthesis system, avoiding theudneed to make accent-specific changes manually. Theudsame linguistic resources can be re-used for each newudaccent. More fundamentally, the keyvowel systemudfunctions as a meta-accent that subsumes vowel-relatedudinformation in all accents of English.