How to design better ontology metrics
Published: 2007 Juni
Herausgeber: Wolfgang May and Michael Kifer
Buchtitel: Proceedings of the 4th European Semantic Web Conference (ESWC'07)
Erscheinungsort: Innsbruck, Austria
Note: to appear
You can only control what you can measure. Measuring ontologies is necessary to evaluate ontologies both during engineering and application. Metrics allow both the fast and simple assessment of an ontology and also to track their subsequent evolution. In the last few years, a growing number of ontology metrics and measures have been suggested and defined. But many of them suffer from a recurring set of problems, most importantly they do not take the semantics of the ontology language properly into account. The work presented here is a principal approach to facilitate the creation of ontology metrics with the clear goal to go beyond structural metrics to proper semantic-aware ontology metrics.
We have developed guidelines and a set of methodological tools based on the notion of "normalization" and "stable metrics" for creating ontology metrics. These guidelines allow the metric author to decide which properties the metrics need to fulfill and to appropriately design the desired metric. A discussion of an exemplary metric (taken from literature) illustrates and motivates the raised issues and suggested solutions.