In other words, for the individual contents whose belonging set class is not identified, the clustering part 253 generates content keyword information having a vector type data structure in which elements are the keywords of the individual contents and a weight is a value based on the score of each of the keywords.