| 摘要 |
PURPOSE:To retrieve a word or phrase only from a text voice without giving a manual work by retrieving the word or phrase in the text voice through the use of the method of a pattern matching from the recognizing phoneme and syllable string of the text voice, the phoneme and syllable string of a retrieval word or phrase. CONSTITUTION:The phoneme and syllable string of the word and phrase, to which the retrieval is desired are stored in a storage means 16, respective phoneme and syllable strings of the text voice and the word and phrase to be retrieved are matched by a matching part 17, and the likelihood of the matching is calculated. By a comparing means 19, based on the calculation result, a part to exceed a likelihood threshold set to a likelihood threshold storage means 18 in the phoneme and syllable of the text voice is extracted, and the phoneme and syllable section of the retrieval result is detected in the text voice. A voice data reproducing start edge pointer 20 indicates the position of voice data from the phoneme and syllable of the retrieval result section and the start edge position of the voice data of a voice data start edge storage means 14, it is converted into the voice by an electric signal/voice converting means 21, and it is outputted. Thus, the retrievals of the word, phrase, etc., can be executed only from the text voice. |
