| Summary: | 碩士 === 國立中央大學 === 資訊管理研究所 === 97 === Search engine has become an essential tool in the era of the information explosion, hence the topic of helping users to filter an excess of information and take personal implicit searching intentions into consideration in order to reach personalized searching ranking has always been important.
Knowledge ontology was used to depict user’s preference and a Chinese keyword recommendation system was proposed to accomplish a Chinese Query Expansion. Analyzing the site maps of the whole user’s past browsing via web crawler, constructing a wider range of personalized domain knowledge automatically by Formal Concept Analysis, and combining Query Expansion and personal ontology which is automatic-learning through HowNet, the more complete information can be accessed easily. When user submits keywords, the system will compare keywords and concepts of personalized ontology in user’s profile in order to produce extended keyword sets similar to the keywords inputted and to be recommended to user to acquire more document information including the same concepts. The experimental results show that the system increases the retrieval precision over 70% and the retrieval precision almost doubles.
By filtering most web documents unconcerned with user’s interests to acquire the actual needed information. The algorithm we proposed that provide automatic-generated user’s knowledge database, a wider range of training data source, a semi-automatic recommended mechanism of Chinese expansion words, and a sememe database of HowNet in Traditional Chinese, is proved to have better retrieval accuracy in the Chinese environment compare to methods of ordinary ontology query expansion.
|
|---|
