From User-contributed Tags to Domain-level Constructs: A Heuristic Approach

The World Wide Web is rapidly becoming the tagged web as citizens tag resources to facilitate retrieval. We argue that these tags capture individual, often incomplete, knowledge about a domain. Aggregated across a large number of contributors, they contain the potential to identify, in a bottom-up manner, key constructs in a domain. This research develops a set of heuristics that aggregates and analyzes web tags to extract domain-level constructs. In addition to inferring their existence, the heuristics distinguish entities, attributes and relationships. This paper presents the heuristics and demonstrates their usefulness with tags from Del.icio.us.