Search Engine Technique Using New Ranking Algorithm for Web Mining

Abstract

Nowadays, there are a huge amount of resources on the Web, which raises a serious problem of accurate search. This is because data in HTML files is useful in some contexts but meaningless under other conditions. In addition, HTML cannot provide description of data encapsulated in it. In many situations a flat list of ten search results is not enough, and the users might desire to have a larger number of search results grouped on-the-fly in folders of similar topics. In addition, the folders should be annotated with meaningful labels for rapid identification of the desired group of results. In other situations, users may have different search goals even when they express them with the same query. In this thesis the research results should be personalized according to the users' on-line activities. There are also situations where users might desire to access fresh information. In these cases, traditional link analysis could not be suitable. In fact, it is possible that there is not enough time to have many links pointing to a recently produced piece of information. In order to address this necessity, we discuss the algorithmic and numerical ideas behind a new ranking algorithm suitable for ranking fresh type of information, such as news articles.