Ontology : a consumer ’ s point of view

Ontology is the science of objects. It is an ancient discipline, which has recently been rediscovered, and overhauled, for the purposes of Artificial Intelligence. Ontology has been concerned with the properties of objects, with their modes of existence and with questions such as how they can be divided in parts and how they fill space. This presentation takes the position of a user of ontologies. It will seek to show not only that there is a ’production of ontologies’ in the research literature, but also that these ontologies are useful and can be used. The paper will concentrate on the latter two issues, focusing on a particular area of application, namely that of Geographic Information Systems (GIS). It will answer the question, how can ontologies be used and how can they contribute to building better information systems? This will mostly be done by showing how the lack of a consistent ontology in a system causes difficulties for its users. The paper will conclude with a set of recommendations as to how ontologies can be made more useful and how the connection between the producers and consumers of an ontology can be structured to make the exchange of ideas more effective. I will also list a number of broad general directions and specific topics which I believe may yield useful contributions both from a scientific and from an engineering point of view. This paper is built on a simple metaphor: ontologies are products and are sold in the international supermarket of AI research. In this supermarket, consumers look for ontologies which fulfill their particular needs. They select ontologies to describe the entities in their application domain and the chosen ontologies must form a consistent ensemble. Unfortunately, contradictions and other inconsistencies may become apparent as one seeks to implement the information system. In consequence, producers of ontologies must put clear labels on the products on the shelf, indicating the advantages and disadvantages of each ontology and indicating also potential conflicts with other ontologies: "Truth in labeling for ontologies!"