| 摘要 |
PROBLEM TO BE SOLVED: To achieve information retrieval matched to a user's taste while improving the accuracy of the information retrieval based on the user taste and reducing the burden of the user, and to obtain efficiently user taste information even by use of a terminal with poor interface performance. SOLUTION: A profile generation part clusters retrieval object information, reconstructs a characteristic space based on the resulting individual clusters, and vectorizes the retrieval object information based on the resulting characteristic space to generate learning data. An information retrieval part uses the learning data for information retrieval. The user is made to select retrieval object information used for user taste information collection from the individual clusters, and considering the result of clustering, minimized retrieval object information such as a musical composition selected from a remote cluster is presented to the user, and taste information for the presented retrieval object information is collected through an evaluation input fetching part. COPYRIGHT: (C)2006,JPO&NCIPI
|
