| 摘要 |
PROBLEM TO BE SOLVED: To provide a speech retrieval device, a speech retrieval method, and a program capable of retrieving a speech with high accuracy and at high speed.SOLUTION: A search character string acquisition unit 111 acquires a search character string. A conversion unit 112 converts the search character string into a phoneme string. A time length derivation unit 113 derives the utterance time length of a speech corresponding to the search character string. A section designation unit 114 designates a likelihood acquisition section in the speech signal to be retrieved. A likelihood acquisition unit 118 acquires likelihood indicating a degree of likelihood that the likelihood acquisition section is a section in which a speech corresponding to the search character string is uttered. A repetition unit 119 changes the likelihood acquisition section designated by the section designation unit 114, and repeats the processes of the section designation unit 114 and likelihood acquisition unit 118. A specification unit 124 specifies, on the basis of the likelihood acquired for each of the likelihood acquisition sections, an estimation section from among the speech signal to be retrieved in which it is estimated that a speech corresponding to the search character string is uttered. |
