| 摘要 |
PROBLEM TO BE SOLVED: To provide a speech recognition device, a speech recognition method and a speech recognition program, which give a suitable recognition result depending on a characteristic of input speech.SOLUTION: The speech recognition device 1 for recognizing input speech based on a language model comprises: a first language model (LM) holding section 130 for holding a first language model created based on a first data base (DB); a first DB length holding section 110 for holding a length of a data of the first DB; a second LM holding section 140 for holding a second language model created based on a second DB; a second DB length holding section 120 for holding a length of a data of the second DB; a weight determination section 170 for calculating a weight based on relation of information regarding a length of an input speech, the length of the first DB and the length of the second DB; an LM mixing section 180 for mixing the first language model and a second language model with the weight; and a recognition processing section 190 for converting the data of the input speech to character, on the basis of the mixed language model. |
