| 摘要 |
A similar document is retrieved by performing a search using diagram information within documents, without being influenced by the description language within documents or the wording of complex sentences. First, feature data (feature amounts) of images is extracted from diagrams that are dotted throughout a document, with respect to a designated document that is designated by a person doing the search. Thereafter, the similarity between documents is evaluated, by comparing the feature amounts of diagrams in the designated document with the feature amounts of diagrams in a document group serving as a search target that are extracted in advance. Ranking of similar documents to the designated document is realized, based on the evaluation result.
|
