Web Search Personalization by User Profiling

The World Wide Web is growing at a rate of about a million pages per day, making it tougher for search engines to extract relevant information for its users. Earlier Search Engines used simple indexing techniques to search for keywords in websites and gave more weightage to pages with higher frequency of keyword occurrences. This technique was easy to trick by using meta-tags liberally, claiming that their page used popular search terms, thereby, made meta-tags useless for search engines. Another technique widely used was to repeatedly use popular search terms in invisible text (white text on a white background) to fool engines. These fallacies called for a set of algorithms which would sort the results using an unbiased parameter. The currently employed Link Analysis Algorithms make use of the structure present in ´hyperlinks´, sorted and displayed depending on a ´popularity index´ decided to pages linking to it. In this work, we have analyzed the mathematics behind these ´link analysis algorithms´ and their effective use in ecommerce applications where they could be used for displaying ´personalized information´.