Stem-ML brings several ideas into intonation modeling:
- we assume that people plan their utterances several syllables
- we assume that people produce speech that is optimized to meet
- we apply a physically reasonable model for the dynamics of the
muscles that control pitch , and
- we introduce a linguistically reasonable concept of a strength
that is associated with each word.
Greg Kochanski, Chilin Shih 2002-08-03