Οικονομικό Πανεπιστήμιο Αθηνών, Τμήμα Πληροφορικής
Web personalization is the process of customizing a Web site to the needs of specificusers, taking advantage of the knowledge acquired from the analysis of theuser’s navigational behaviour (usage data) in correlation with other informationcollected in Web context, namely structure, content and user profile data. ISPshave full knowledge of the visited Web graph by all the subscribers’ navigationalpatterns and all users’ demographic and context related data; as such, they can addrecommended advertisements in every web page the user visits in real time. Thisgives a potential for an ISP Advertising framework that is targeted. The proposedsystem for personalization on the web will be implemented in two parts, the firstsituated on the ISP’s server and the other on the router provided to the subscriber;the latter is the subject of this thesis. The system offers the ability to provide realtime personalization services to customers. The system will collect informationconcerning the navigation and preferences of users as they browse the Internet. Byutilizing usage-data, data mining techniques and learning procedures, user profileswill be created as well as appropriate sets of recommendations that will be relatedto both their long and short term interests. By extracting keywords from the currentweb page and scoring them according to a local dynamic term corpus, a webpage vector representing the current web page is constructed. In similar fashion, auser profile vector is maintained and updated as the user browses more web pages.Both these vectors are transmitted to the ISP’s server in order to find the most relevantadvertisements. The recommended advertisements will be introduced in realtime into the web page content as traffic is passing through telecommunicationdevices (such as a router) situated close to the user.
Internet
Information retrieval
Computer science