| 摘要 |
<P>PROBLEM TO BE SOLVED: To provide a sentence classification apparatus capable of highly accurately classifying sentences obtained from a call. <P>SOLUTION: A sentence classification method classifies a sentence which is a speech recognition result of voice data of a calling voice into first sort or a second sort. The method includes steps (S1000 to S1100) for input of a sentence and steps (S1200 to S2100) for finding out a probability value indicating probability that the input sentence is the first sort from a value obtained by dividing the appearance frequency of a phrase included in the input sentence in sentences of the first sort by an added value of the appearance frequency of the phrase in the sentences of the first sort and the appearance frequency of the phrase in sentences of the second sort and then classifying the input sentence based on the found probability value. In the classification steps, the probability value is corrected so that influence exerted on the probability value is more suppressed in phrases with lower reliability as speech recognition results(S1500). <P>COPYRIGHT: (C)2011,JPO&INPIT |
