Both topic maps specifications and literature have implicitly or explicitly presented the standard as ‘ontology-agnostic’, meaning they are able to support, represent and manage any kind of knowledge in any kind of ontological context, and even independently of the constraints imposed by any ontology. Topic maps are able to express ‘anything about anything whatsoever’. This very claim has attracted some criticisms from experts in formal ontologies, and among the Semantic Web community, sometimes interpreting it like a puff of smoke hiding poor formal foundation.

Beyond theoretical claims and academic debate, most if not all real-world topic map implementations and software use explicitly or implicitly schemas and/or ontologies, defining classes of topics, role types and association templates, occurrence type constraints and data structure… Such sets of constraints should eventually be expressed as topic map schemas, but the relevant specification, TMCL, based on Topic Map Data Model, has still a long way to go.

Meanwhile, rapid development and adoption of ontologies in a wide range of domains and industries, and singularly based on the new W3C Web Ontology Language OWL, have put the issue in a slightly new perspective. If OWL ontologies are to become effectively mainstream core technology for the Semantic Web, topic maps applications cannot afford any more sticking to the agnostic line, pretending to play in a somehow orthogonal field.

In fact topic maps would indeed gain effectiveness and interoperability either through explicit formalization of ontologies specifically built and dedicated for topic map control, or through declaration of commitment to pre-defined ontologies, not specifically designed for that use. In either case, using OWL ontologies should be considered as the most interesting choice.

Based on examples chosen from actual Mondeca ITM implementations of ontology-driven topic maps, the presentation will show the benefits of this approach, and their current limits.