Title:
主題抽出装置およびそのプログラム
Document Type and Number:
Japanese Patent JP5856905
Kind Code:
B2
Abstract:
PROBLEM TO BE SOLVED: To provide a subject extraction device capable of highly accurately carrying out weighting of words by using a situation where a word appears and inter-word semantic relation.SOLUTION: The subject extraction device comprises: an expression extracting unit for extracting expressions from read document data; a relation acquiring unit for acquiring relation data which shows semantic relation between expressions; a first importance evaluating unit for calculating first importance based on an occurence frequency of the extracted expression in the document data; and a second importance evaluating unit for, with two expressions included in the document data defined as a start point expression and an end point expression respectively, calculating second importance of the end point expression on the basis of the sum of relation density from the start point expression to the end point expression when the start point expression and the end point expression are connected by one or more stages of relation by using the relation data acquired by the relation acquiring unit, and calculating total importance based on the first importance and the second importance about the expressions included in the document data.
Inventors:
Ichiro Yamada
Masaru Miyazaki
Hideki Tanaka
Masaru Miyazaki
Hideki Tanaka
Application Number:
JP2012116759A
Publication Date:
February 10, 2016
Filing Date:
May 22, 2012
Export Citation:
Assignee:
Japan Broadcasting Corporation
International Classes:
G06F17/30; G06F17/27
Domestic Patent References:
| JP2011086043A |
Other References:
中嶋卓雄,他,“ニュースフィルタリングエージェントにおける協調”,マルチメディア通信と分散処理ワークショップ論文集,日本,社団法人情報処理学会,1995年10月25日,第95巻,第2号,p.63-69
Attorney, Agent or Firm:
Masatake Shiga
Tadashi Takahashi
Tadashi Takahashi