Document Topic Detection Based On Semantic Feature

• Main Authors: Shu-Chuan Li, 李淑娟
• Other Authors: Hei-Chia Wang
• Format: Others
• Language: zh-TW
• Published: 2007
• Online Access: http://ndltd.ncl.edu.tw/handle/90689058555050318781

碩士 === 國立成功大學 === 資訊管理研究所 === 95 === Because of the development of Internet, electronic journals have become a trend. The quantities of people use electronic journals more than before. The amount of electronic journal articles grow faster than before, it leads that information generated over the ability that people can deal with. In order to deal with this problem, a lot of electronic periodical databases have proposed keyword search methods to lighten user's effort and time spending in searching the journal papers. However, the users still have to face the huge search results. Currently, there is no better way of representing to help users to speed up filtering relevant documents. How to provide a efficient search, i.e. present the search result in categories, have became an important research topic now. Today’s electronic journal databases apply keywords search which user inputs interested terms and search engine find the papers which contains keywords and show the searching results by the way of tabulating. If these results can be generalized and classify by their topics, then we can show the search results to users by the topic which should be able to improve tradition display method. Though scholars have employed topic detection method to achieve this goal in full-text documents on network. However, literatures have the structure properties, such as title, keyword, and abstract, not only full-text. Simultaneously, traditional topic detection method only uses the word frequency feature, ignores the importance of semantic. Therefore, the proposed research designs a method which is based on literature structure and semantic properties to extract important words and cluster to each literature. It can retrieve each group topics and display the searching results by these topics. Expect users can reduce literature collection time and find correctly information by topic-cluster display way.