SIMILARITY INDEX VALUE CALCULATION DEVICE, SIMILARITY SEARCH DEVICE AND PROGRAM FOR CALCULATING SIMILARITY INDEX VALUE

摘要

To provide a similarity index value calculation device, a similarity search device, and a program for similarity index value calculation capable of improving evaluation accuracy of similarity than conventional ones.SOLUTION: A similarity index value calculation device includes: a word extraction unit 11 that analyzes m sentences and extracts n words; a vector calculation unit 12 that calculates, by vectorizing the m sentences and the n words into a q-dimensional space, m sentence vectors consisting of q axis components, and the n word vectors consisting of q axis components; and an index value calculation unit 13 that calculates, by respectively taking inner products of the m sentence vectors and the n word vectors, similarity index values each reflecting a relationship between each of the m sentences and each of the n words. Thus, the similarity index value calculation device enables obtaining the similarity evaluation values representing, as inner product values, which word contributes to what sentence to which degree, by calculating the inner products of the sentence vectors calculated from the sentences and the word vectors calculated from the words included in each sentence.SELECTED DRAWING: Figure 1