Automatic query generation based on user bookmark file folders

Abstract

The World Wide Web is a growing base of pages with large amounts of information in it. Users who are browsing through these Web pages will undoubtedly find certain pages which they would like to revisit in the future. Bookmarks and Favourites store the URLs of those pages for the user so that he can easily return to them later. This will make internet browsing easier for the user. However these Bookmark managers offer little features other than storing the URLs. While searching is playing a major role in the internet, there have been no searching options within the three most used Web browsers bookmark managers. Hence, this thesis presents two algorithms which provide Web site recommendations based on a particular bookmark category. These two different algorithms will process the categories found in the bookmarks files and generate queries which are related to the URLs found in the category. The first algorithm will get words from the URLs found stored in the bookmark category. The second algorithm will find sites which link to the bookmarked URL and gets the words from them and from the stored bookmark. These algorithms are evaluated using an evaluation platform. In the first part of the evaluation, the user is allowed to construct his own query with the aim to find Web sites similar to the ones in the bookmark category. After giving his feedback, the user is then given the results of the automatically generated query to give feedback on them. Results from the evaluation carried out to test the relevance of the queries showed quite promising results. User generated queries gave 89.9% relevant results with 53.2% of the relevant results being good enough to bookmarked. Automatically generated queries gave 77.9% relevant results with 53. 7% of the relevant results being good enough to bookmark.