Personalized meta-search system

Abstract

The size of the Internet is growing dramatically every second. A general search engine could only capture a limited part of the Internet. A meta-search engine could be a solution to the above, by aggregating search results from both general and specialized search engines. User could research for topics they want in the huge Internet from a single location. In this project, a console based meta-search engine was developed. Consolidated results from both private/public resources will be available for user to review upon their request. Coverage of all the search engines could be merged together. Various components of the meta-search Engine will has been studied, includes, Search Engine agents, Automatic Search Engines selector, and Result Merger. Where advanced gears of information retrieval have been made used for example the Vector Space Model, Naive Bayesian Classifier. The solution provide a unique solution compare to other researches, where besides ranking from search engine, others readable information is also considered. The summaries, the tiles and the actually content would all be read by the system. Therefore to make the system more like a real human on reading and ranking for the user. In order to build a more effective search engine, personalization by learning algorithm were implemented into the project. Where user could suggest what they want, the system will then prepare it for the user. The experience is accumulated to target the purpose of personalization, where the experience would be centralised into a database. On the other hand, statistical method is used to estimate the user preference towards the search engines concerns. In more specific, Linear Regression model would be used to correlate the preference of search engines to the user. A complete framework was be studied and developed. Due the limitation in times the algorithm combination may not be good to fully evaluate user preference; the project could have some implication on what a meta-search engine should be or what a meta-search engine should not be.