摘要 |
PROBLEM TO BE SOLVED: To solve the problem that when high speed processing is performed by clustering in similarity retrieval intended for high dimensional vector data, it is not generally guaranteed that it is matched with retrieval results acquired by referring to whole cases, and when high precision is requested, it is necessary to refer to much more clusters, and it is not possible to achieve any quick response. SOLUTION: Information relating to retrieval results and referenced clusters is transmitted and received between a retrieval engine and an application under such conditions that the number of clusters to be referenced by the retrieval engine one time is fixed. The retrieval engine omits processing relating to the already referenced clusters, and executes processing relating to the fixed number of clusters. Thus, it is possible to configure more precise retrieval results in a fixed response time. COPYRIGHT: (C)2008,JPO&INPIT
|