Augmenting Web Page Classifiers With Social Annotations

The lack of representative textual content in many web documents suggests the study of additional metadata to improve web page classification tasks. Social bookmarking sites provide an accessible way to increase available metadata in large amounts with user-provided annotations. This field remains relatively unexplored. In this work, we analyze the usefulness of social annotations for web page classification. We evaluate the results on two different categorization levels, and analyze their suitability for home and deeper pages. We conclude that social annotations could enhance web page classifiers in multiple cases, and we present a method to get the most out of them using classifier committees.