| 摘要 |
PROBLEM TO BE SOLVED: To extract a proper semantic label according to a content without previously selecting a semantic label to be imparted to accurately generate an image dictionary. SOLUTION: Content related information that is text information wherein contents of the video content are explained is acquired from a Web site, the text information related to the content is collected from the Web site based on a content disclosure data or a content creation date, and a word extracted from the content related information, a plurality of words expressing a topic are extracted from the text information as a word set, a Web image related to the word set is collected, a visual pattern related to the topic is generated with the collected Web image as learning data, and the word set and the visual pattern are stored in an image dictionary storage means as the image dictionary. COPYRIGHT: (C)2008,JPO&INPIT
|
