| 摘要 |
PROBLEM TO BE SOLVED: To provide an audio information retrieval device which can retrieve the information in voice at a high speed and with a high efficiency without forming a concept network where the inter-language relation is previously defined. SOLUTION: This retrieval device recognizes input voices, generates word candidate series through the collation processing with a word dictionary 2, and compares the word index of a word appearance information storage part 6 that is extracted from an example sentence storage part 3 with the word index of the said word candidate series to calculate the evaluation point against the corresponding example sentence. Then the word index of the optimum candidate word series and also the optimum example sentence are selected. Thus, it is possible to select the optimum example sentence and also the information and therefore to retrieve the information with high efficiency and high accuracy. |
