| 摘要 |
<p>PROBLEM TO BE SOLVED: To automatically detect that a speaker has repeatedly uttered words and phrases classified as a same item, by adding to the utterance a language component code of a code column composing a speech recognition data, and providing a overlapping judging means, etc., for judging from the summed value whether or not the language of the same item is overlapped. SOLUTION: In a speech recognizing part 30, a speech recognizing means 31 sequentially recognizes from speech signals words and phrases each of a manufacturer's name, a product name, and the number of articles purchased in this order. Language component codes defined corresponding to each of the words and phases are sequentially extracted from speech recognition resources by a language component extracting means 32. Speech recognition data are generated by a speech recognition data creating means 33 from a code string of the language component codes corresponding to each of the words and phrases extracted from these speech recognition resources 20. After that, an overlapping judging means 34 judges whether or not languages of a same item overlap in the speech recognition data generated by the speech recognition data creating means 33.</p> |
