| 摘要 |
<p>An automatic inquiry system includes a storage (52) for storing a plurality of inquiry entries, each entry comprising a plurality of data fields. For at least one data field of each inquiry entry a respective reference transcription set is stored which includes at least two alternative acoustic transcriptions of data stored in the data field. A dialogue engine (50) executes a human-machine dialogue to determine a plurality of pre-determined query fields. Instead of fully recognizing all utterances, a speech recognizer (40) is able to represent at least one utterance as a corresponding input transcription set, which includes at least two alternative acoustic transcriptions of the utterance. The dialogue engine (50) specifies at least a first one of the query fields by an input transcription set of a selected one of the utterances. A search engine (59) locates inquiry entries in the storage whose associated reference transcription set relates to the first one of the query fields. In this way, utterances, such as specifying family names, which are difficult to recognize accurately need not be fully recognized before a query is made.</p> |
