Automatic and adaptive iNotify web candidate web page identification

Abstract

During the last years, the amount of information found on the internet has increased at a high rate, making web-browsers one of the most used modem-day interfaces. Apart from this increase in information, the internet has developed from a static to a dynamic one, continuously delivering up-to-date information. Despite of all the developments of the internet-browsing experience, the web-browsers' interfaces have not yet developed accordingly and do not provide suitable re-visitation tools. Up to this date the user still encounters such problems as the Cognitive Overload problem and the Lost in Hyperspace problem. The system presented in this dissertation is such that automatically and adaptively identifies candidate web-pages for re-visitation and monitors them for updated content. Web browsing revisitation features and web browser history are retrieved from a Firefox extention and sent to an ASP .NET website for further processing. A decision tree is implemented to classify each web history instances as either "interesting" or "not interesting". A vector space model with history session information is implemented to classify the "interesting" web-pages into categories "User Interests". Comparison between web page cache versions and live versions of DOM trees are used to identify the updated content from the candidate web pages. Results are displayed to the user, relieving him/her from checking the updated content of a web site. Results show that web-page re-visitation features and web history aid in automatically and adaptively identify candidate web-pages for re-visitation.