| 摘要 |
It is possible to automatically extract information to be considered from a plenty of information. A disclosure unit of a personal Web site and a person al opinion on a bulletin board is extracted from the collected content information and information (URL and opinion number) for specifying the disclosure unit of the personal opinion is registered. Next, an object (company name, category of business) of the personal opinion is identified. By analyzing the disclosed content of the personal opinion, a personal evaluati on (good evaluation/bad evaluation) of the object is identified. Moreover, reliability degree is decided according to whether a referenced ranking, an opinion basis, information indicating a speaker~s identity are contained. Thus, it is possible to indicate the evaluation of the object as a personal opinion. Moreover, it is possible to extract, for example, only band evaluations among the evaluations of the objects as personal opinions. Moreover, it is possible to search opinions having a large influence accordi ng to the referenced ranking and reliability degree.
|
