摘要 |
PURPOSE: A speech recognition apparatus having a language model adaptive function and a method for controlling the same are provided to improve the speech recognition rate and a response characteristic by revising a language model. CONSTITUTION: An acoustic model element(14) constructs an acoustic model related to speaking data for learning. A pronunciation lexicon(22) mentions words to be pronounced. A sentence data unit(16) is constructed with sentences for learning. A frequency data unit(18) has frequency data related to word sequence of the sentences recorded on the sentence data unit. A probability data unit(20) has probability data related to the calculated word sequence by using the frequency data. A recognition processing unit(26) recognizes inputted speech for combining the words of the pronunciation lexicon for generating a recognized sentence. An adaptation sentence processing unit(28) inputs the sentence received from the user and a weight value of the word sequence in the sentence, and revises the frequency data and the probability data.
|