| 摘要 |
<P>PROBLEM TO BE SOLVED: To prevent deterioration in pronunciation presuming accuracy due to accidental co-occurrence of a wrong pronunciation hypothesis and written expression of a word that is an object of pronunciation presuming and to achieve highly accurate pronunciation presuming. <P>SOLUTION: A pronunciation hypothesis generating unit 101 generates a plurality of pronunciation hypotheses for a word whose pronunciation is to be presumed; a co-occurrence score calculating unit 103 figures out, with respect to each of the plurality of pronunciation hypotheses, a co-occurrence score by using a co-occurring relationship between each hypothesis and the word whose pronunciation is to be presumed in a group of documents prepared in advance; a hypothesis selecting unit 105, on the basis of the co-occurrence score, selects one or more pronunciation hypotheses out of the plurality of pronunciation hypotheses; here the co-occurrence score calculating unit 103 figures out not only the number of documents in which the word whose pronunciation is to be presumed and the pronunciation hypotheses occur but also the distance between the word whose pronunciation is to be presumed and the pronunciation hypotheses in the document as the co-occurring relationship, and calculates the co-occurrence score on the basis of the co-occurring relationship so figured out. <P>COPYRIGHT: (C)2012,JPO&INPIT |
