| 摘要 |
In the disclosed speech recognition based interactive information retrieval scheme, the recognition target words in the speech recognition database are divided into prioritized recognition target words that constitute a number of data that can be processed by the speech recognition processing in the prescribed processing time and that have relatively higher importance levels based on statistical information, and the other non-prioritized recognition target words. Then, the speech recognition processing for the speech input with respect to the prioritized recognition target words is carried out at higher priority, and a confirmation process is carried out when the recognition result satisfies a prescribed condition for judging that the retrieval key can be determined only by a confirmation process with the user. On the other hand, a related information query to request the user to enter another speech input for a related information of the retrieval key is carried out when the recognition result does not satisfy the prescribed condition, and the original recognition result is adjusted according to the recognition result for another speech input. In this way, the retrieval key determination is realized through natural speech dialogues with the user.
|
