Ordinary Web pages as a source for metadata acquisition for open corpus user modeling

Abstract

Personalization and adaptivity of the Web as we know of today is often “closed ” within a particular web-based system. As a result there are only a few “personalized islands ” within the whole Web. Spreading the personalization to the whole Web either via an enhanced proxy server or using an agent residing on a client-side brings a challenge how to determine metadata within an open corpus Web domain, which would allow for an efficient creation of overlayed user model. In this paper we present our approach to metadata acquisition for open corpus user modeling applicable on the “wild ” Web, where we decided to take into account metadata in the form of keywords representing the visited web pages. We present the user modeling process (which is thus keyword-based) built on the top of an enhanced proxy server, capable of personalizing user browsing sessions via pluggable modules. The paper focuses on comparison of algorithms and thirdparty services which allow for extraction of required keywords from ordinary web pages, which is a crucial step of our user modeling approach.