Words to Phonemes

This phase converts the words in a word sequence into a phoneme sequence. The phoneme sequence is internally represented using a feature-based phonetic model similar to that specified in Kirshenbaum’s ASCII-IPA document. This allows different phoneme transcription schemes to be used for the language and voice.

The phonemes are generated using the specified language (either from the document or explicitly specified by the user) using letter-to-phoneme rules and an exception dictionary. The letter-to-phoneme rules describe how letter sequences are pronounced and the exception dictionary provides a set of words that are not pronounced correctly with the letter-to-phoneme rules.


  1. Phoneme Transcription Schemes.
  2. Rule Engine.