SPEECH RECOGNITION SYSTEM AND METHOD

摘要

PURPOSE: A voice recognition system and a method thereof are provided so that the change of utterance is considered. The voice recognition performance is improved. CONSTITUTION: An estimating unit(110) presumes the language model. The dictionary(130) stores the corresponding information between phoneme series expressing the phoneme series of the notation of word, and utterance change and the word. The recognition graph preparation part(140) writes out the recognition graph based on the presumed language model as described above and the corresponding information about the word included in the language model.