PROGRAM, DEVICE, AND METHOD FOR CALCULATING SIMILARITY BETWEEN CONTENTS REPRESENTED BY SETS OF FEATURE VECTORS

PROBLEM TO BE SOLVED: To provide a program, device, and method which are capable of accurately calculating similarity between a query content and an object reference content.SOLUTION: The device includes: reference feature vector search means which acquires a set of reference feature vectors from reference feature vector storage means with respect to each query feature vector of a query content and searches for N (N>1) reference feature vectors in the descending order of similarity; object reference feature vector search means which searches for an object reference feature vector having the highest similarity in an object reference content; score calculation means which calculates scores on the basis of a ranking showing the position where a distance between the query feature vector and the object reference feature vector having the highest similarity is placed in ranking of distances between the query feature vector and the N reference feature vectors; and similarity calculation means which calculates, as similarity, the sum total of scores of all query feature vectors.