| 摘要 |
<P>PROBLEM TO BE SOLVED: To detect a difference and a change in the concept of a word produced among concept dictionaries, caused by different document data groups in regard to text mining. <P>SOLUTION: A coincidence frequency calculation section 102 calculates a coincidence frequency having coincident occurrence of a combination words from a first document data group and a second document data group, so as to set as a first coincidence frequency table and a second coincidence frequency table. A singular value solution section 104 performs singular value solution of the first coincidence frequency table and the second coincidence frequency table, so as to obtain a first concept dictionary and a second concept dictionary constituted of concept vectors on a word-by-word basis in a signification space of reduced dimension. From a concept vector for each word in the first concept dictionary and the second concept dictionary, a concept dictionary difference extraction section 109 calculates a difference of a word to be compared from each criterion word, and obtains a difference word of the concept dictionaries, thus obtaining concept dictionary difference information. <P>COPYRIGHT: (C)2006,JPO&NCIPI |
