Adaptive automatic classification on the Web

The paper explores the motives and issues associated with automatic classification on the Web. Experimental software tools are introduced which enable the automatic generation of vocabularies, representing any given classification scheme, using pre-classified documents. An internal representation of the required classification hierarchy is automatically generated and subsequent documents are filtered through this hierarchy and automatically classified. The flexible, dynamic manner in which the classification vocabularies are defined creates an environment in which different collections of documents can be classified according to different, extensible classification schemes. Evolving changing vocabularies and hierarchies can be dynamically generated and modified over time.