Improving Search in Tag-Based Systems with Automatically Extracted Keywords

Abstract

Tag-based systems are used by millions of web users to tag, save and share items. User-defined tags, however, are so variable in quality that searching on these tags alone is unsatisfactory. One way to improve search in bookmarking systems is by adding more metadata to the user-created tags to enhance tag quality. The additional metadata we have used is based on document content and largely avoids the idiosyncratic and ambiguous terms too often evident in user-created tags. Such an approach adds value by incorporating information about the content of the resource while retaining the original user-created tags.

This paper describes how users’ tags can be enhanced with metadata automatically extracted from the original document. An experiment comparing search based only on user-created tags with search using an automatically enhanced tag set, demonstrates how incorporating the extra tags can offer significant benefits.