Title:
INFORMATION RETRIEVAL DEVICE
Document Type and Number:
Japanese Patent JP3632359
Kind Code:
B2
Abstract:
PROBLEM TO BE SOLVED: To attain the clustering of a large number of pieces of document information with small computational complexity by selecting a prescribed number of pieces of information as the leaf nodes of a cluster out of plural pieces of information and allocating the other information to the similar leaf nodes.
SOLUTION: At a control part 123 of an information retrieval server 12 connected to an internet 11, a leaf node information selection means 123a selects a prescribed number of optimum information out of plural pieces of page information. A partial cluster generation means 123b allocates other page information to the similar leaf nodes of a cluster to generate a new cluster. Then a recursive clustering means 123C designates both means 123a and 123b in the leaf node direction of the generated cluster so as to repetitively operate them again. Therefore, the accuracy of similarity of clusters is decided based on the information selected by the means 123a and a cluster is generated at a higher speed.
More Like This:
Inventors:
Keiko Aoki
Kazunori Matsumoto
Kazuo Hashimoto
Kazunori Matsumoto
Kazuo Hashimoto
Application Number:
JP6749697A
Publication Date:
March 23, 2005
Filing Date:
March 06, 1997
Export Citation:
Assignee:
KDDI CORPORATION
International Classes:
G06F17/30; (IPC1-7): G06F17/30
Other References:
姚左軍、濱田喬,類似検索における特徴ベクトルのインデックス及び関連の検索に関する一手法,情報処理学会論文誌,日本,社団法人情報処理学会,1996年11月15日,37巻、11号,2056-2064頁
Attorney, Agent or Firm:
Keiichi Yamamoto