BBC Broadcasters Select HTK for Database

HTK has added its speech synthesis system to an existing BBC database, creating a computer generated voice which accurately pronounces foreign words and names in English Broadcasting context so presenters can hear how to say them correctly before going live on air.

 

HTK worked closely with BBC's in-house pronunciation research team to adapt the database, which has been compiled over 80 years of broadcasting and includes over 200,000 pronunciations. The original database relied on modified spelling, a technical language that shows each preferred pronunciation as a sequence of individual sounds and intonations in a simple format, much like a spelling. HTK has enabled the system with a spoken output that renders the pronunciation, so anyone with a networked computer can not only see the "spelling" but hear it as well.

 

The speech synthesis capability used in the system is based upon a Laureate Text-to-Speech (TTS) Synthesizer maintained by HTK under licence from BT. Laureate provided a starting point for this project, as it works with small pre-recorded units of sound (the vowels, consonants and other elements of speech that make up a language), allowing foreign sounds that don't occur in typical UK English to be recorded and added to the system by HTK.