摘要 |
<P>PROBLEM TO BE SOLVED: To generate description on a speech document by which an influence on the retrieval accuracy of an error in speech recognition is reduced and retrieval by a key associated with a topic in the speech document is retrieved. <P>SOLUTION: In the speech document retrieval method, the speech document is generated by applying a speech recognition processing to the inputted speech document, acquiring a literate speech recognition result together with recognition reliability, collating with expansion key word extract conditions and the speech recognition result to extract an expansion key word, collating with the extracted expansion key word and the expansion word extract conditions, retrieving an external database based on the resulting acquired conditions, extracting the expansion word, and embedding the expansion word in the results of the expansion word and the speech recognition. <P>COPYRIGHT: (C)2004,JPO&NCIPI |