Agreeing to disagree: Reconciling conflicting taxonomic views using a logic-based approach

Taxonomy alignment is a way to integrate two or more taxonomies. Semantic interoperability among datasets, information systems and knowledge bases is facilitated by combining the different input taxonomies into merged taxonomies that reconcile apparent differences or conflicts. We show how alignment problems can be solved with a logic-based region connection calculus (RCC-5) approach, using five base relations to compare concepts: congruence, inclusion, inverse inclusion, overlap and disjointness. To illustrate this method, we use different geo-taxonomies, which organize the United States into several, apparently conflicting, geospatial hierarchies. For example, we align T_CEN, a taxonomy derived from the Census Bureau's regions map, with T_NDC, from the National Diversity Council (NDC), and with T_TZ, a taxonomy capturing the U.S. time zones. Using these case studies, we show how this logic-based approach can reconcile conflicts between taxonomies. We have implemented these case studies with an open source tool called Euler/X which has been applied primarily for solving complex alignment problems in biological classification. In this paper, we demonstrate the feasibility and broad applicability of this approach to other domains and alignment problems in support of semantic interoperability.