Research And Application Of Specialized Website Classification Based On Self-studying

The thesis studies some key technologies in classifying specialized websites. A self-study classification system is constructed on the base of the research, with the capability to retrieve all the websites of a specialized field. The retrieval starts with a small set of seed websites, which provide for the system a profile of this field represented by a keyword model. The expansion of this specialized website set is done by following the hyperlinks on the pages of the current sites. Hyperlink relevance and website hierarchy are analyzed in detail to ensure only most relevant part of the links and pages are processed, avoiding the situation of straying away in the enormous amount of links and pages.The research done in the thesis highlights the following three areas: a keyword model representing the theme of a set of websites; the distribution of relevant links leading to a possibly new website of the specialized field; and page-theme relevance analysis aiming at finding the a small set of webpages of a site that can best represent the theme of the website.The key words come from the "seed" website set,and these key words could be adjusted when the set expands. The appropriate keyword model includes Vector Space Model, Probabilistic Model.A website's pages are organized in hierarchy, including homepage, the second-level pages, the third-level pages and so on. By analyzing the website's hierarchy, quantitive relation between the website's theme and its hierarchy are made out. Then, we can get the pages that could represent the website's theme, the construction of the key word model and the theme evaluation are both based on these pages.