| 摘要 |
PROBLEM TO BE SOLVED: To acquire more detailed and more correct information about a retrieval target word. SOLUTION: Documents including person's names name are extracted from a document set U of documents (web page) (S11), and elements of a document set S which belong to the same URL host group are put together with one another to form a set H of web pages (S12). Next, contents of web pages belonging to respective sets H are subjected to morphological analysis, and names other than the person's names name are extracted (S13). Then, a link with weight r corresponding to a degree of association is generated with a set of web pages for "workspace" ws as a node to prepare a graph G (S14). Nodes with one another are classified into seeds according to a degree of association between nodes from the graph G (S15 and S16), and in addition, even remote nodes are made to belong to a seed with the highest degree of association (S18). COPYRIGHT: (C)2006,JPO&NCIPI
|
