FREELY SPOKEN SPEECH RECOGNITION DEVICE

摘要

PURPOSE: To provide the freely spoken speech recognition device which can automatically generates a statistical language model in consideration of the probability of redundant words and obtains a higher speech recognition rate of free speaking than a conventional example. CONSTITUTION: The freely spoken speech recognition device recognizes the speech of an inputted freely spoken speech sentence consisting of a word string by referring to a specific statistical language model; and a statistical language model stored in a memory 7 is determined and generated (language model generation part 10) with 1st occurrence probability corresponding to a string of basic words that do not belong to a specific predetermined category of word strings and 2nd occurrence probability corresponding to a group of >=0 string of inserted words as words that are inserted between words of basic words including the part before the 1st word of the string of the basic words and the part behind the final word when the string of basic words is given and belongs to the specific category. When the basic words succeed, it is considered that zero word is interposed between them.