| 摘要 |
<p><P>PROBLEM TO BE SOLVED: To provide a coincidence matrix generation device for generating coincidence vectors or conceptual vectors on which semantic similarity and identifiability between corresponding words can be precisely reflected in a method based on inter-word and component numbers. <P>SOLUTION: A coincidence matrix generation device includes: a first clustering means 11 which inputs a first coincidence matrix 14 in which respective lines are made to correspond to words, and respective columns are made to correspond to N pieces of component numbers, and clusters the group of the line vectors of the first coincidence matrix 14 into N' pieces of clusters, and associates the component numbers of the words and clusters; a second coincidence matrix generation means 12 for generating a second coincidence matrix 17 in which respective lines are made to correspond to the words of a text whose morphemic analysis has been performed, and respective lines are made to correspond to N' pieces of component numbers; and a third coincidence matrix generation means 13 for generating a third coincidence matrix 18 in which values obtained by linearly connecting the corresponding elements of the first coincidence matrix 14 and the second coincidence matrix 17 with respect to the optional words and the optional component numbers are defined as corresponding elements. <P>COPYRIGHT: (C)2011,JPO&INPIT</p> |
