Personalized Ranking for Meta-Search Engine by Using SSVM
| Autor: | Kuan-Chen Chin, 金冠辰 |
|---|---|
| Rok vydání: | 2005 |
| Druh dokumentu: | 學位論文 ; thesis |
| Popis: | 93 With the fast growth of the World Wide Web, the amount of information on the Web has become overwhelming. Since current search engines are built to serve all users, the search results are usually ranked based on the public interests of the users. Furthermore, current search engines always provide the same results for all users no matter which field the user belongs to. Hence, some users have to browse the search results laboriously to find out the desired web pages. In this thesis, we build up a personalized meta-search engine (PMSE) which allows users to rank web pages according to their personal interests. A user's interests are represented in a user profile which can be learned from the documents that are stored in the personal computer. Due to the superior performance of support vector machines (SVMs), smooth support vector machine and several other text-classification techniques are applied to obtain user profiles. We take advantage of meta-search engine's better coverage of the Web to collect a wide variety of web pages. Meta-search engine works as a data collector, and two basic meta-ranking algorithms are proposed in our system. For personalized ranking, we re-rank the collected web pages by consulting with the user profile and the search engine preference mechanism (SEP). We simulate 6 scenarios in order to evaluate the performance of our system, which conventional search engines could not provide satisfied search results for the users. Our experimental results indicate that we successfully provide a way for the users to rank the web pages according to their interests, and also show that the personalized ranking is worth pursuing further. |
| Databáze: | Networked Digital Library of Theses & Dissertations |
| Externí odkaz: |
|