Measuring Long-Term Ontology Quality: A Case Study From the Automotive Industry

The use of ontologies based on knowledge representation architectures to support search and other decision-making problems in production environments has become a critical component of information systems. The process of building such an ontology can now take advantage of tools such as Protégé (Gennari et al. 2002) to build an ontology for any given problem domain. There has also been corresponding work done on the development of tools and utilities that measure the "quality" of an ontology and the metrics that can be used to measure different facets of the ontology. In this paper we analyze an existing ontology that has been in use for fifteen years in the domain of process planning for automotive assembly. This system, originally known as the Direct Labor Management System (DLMS), was developed and deployed at Ford Vehicle Operations in the early 1990s (Rychtyckyj 1999). The requirement for maintaining the DLMS knowledge base over the last fifteen plus years has given us a unique perspective into the various maintenance problems and issues that need to be addressed. This paper will discuss those issues and try to frame the ontology quality issue in terms of our experience at Ford Motor Company.