The knowledge base provides the information the decoder needs to do its job. Typically, the knowledge base consists of the acoustic model, the language model, and the lexicon.
KB.gif
The Acoustic Model provides the knowledge for converting frame sequences into unit hypotheses, the Lexicon provides the pronunciation (unit sequence) and part−of−speech classification for words, and the Language Model provides the knowledge for converting unit sequences into word and word sequence hypotheses.

Last edited Oct 12, 2014 at 6:24 PM by DxN, version 7