| 摘要 |
<P>PROBLEM TO BE SOLVED: To provide a causal word pair extraction device of easily extracting a highly accurate domain-unique causal word pair from a document group. <P>SOLUTION: Problem/cause expression relevance calculation means 85 calculates relevance between extracted problem expressions and cause expressions on the basis of the likelihood score of each causal word pair candidate included in the set of the problem expressions and the cause expressions, the feature degree of the problem word of the causal word pair candidate in the problem expressions, and the feature degree of the cause word of the causal word pair candidate in the cause expressions. When a difference between the relevance of the problem expression/cause expression pair calculated by the problem/cause expression relevance calculation means 85 and a relevance between problem expressions in the problem expression/cause expression pair and cause expressions different from cause expressions in the problem expression/cause expression pair is equal to or less than a predetermined threshold, causal word pair learning means 86 changes the likelihood score of the causal word pair candidate used for the calculation of the relevance in accordance with a value which is proportional to the product of the feature degree of the problem word in the problem expressions and the feature degree of the cause word in the cause expressions. <P>COPYRIGHT: (C)2013,JPO&INPIT |
