DeMoSt: a Tool for Exploring the Decomposition and the Modular Structure of OWL Ontologies

Motivation In relevant application fields ontologies are often maintained as huge monolithic collections of axioms in single files, as for some ontologies in the NCBO BioPortal repository4, like the Gene Ontology (∼ 60, 000 axioms). Such representation is not ideal for applications which only require access limited to individual fragments of the ontology. As an example, the reasoning service provided needs just a small part of it to be performed, and loading only the necessary bit to the task can reduce substantially time and memory required, improving the performance of the system. Hence it is important to investigate the possibility of maintaining ontologies in a more flexible form. Recently the notion of module has been developed. Modules are subsets of ontologies that preserve all entailments over a set of terms Σ called seed signature, and they come in many flavours. We focus on modules based on syntactic locality [2] (from now on simply called modules), that are efficiently computable. Modules are useful and used, for example in reuse scenarios, i.e. when a part of a well-established ontology is selected to be reused in a different ontology. However, it is really important to exactly know the seed signature needed, or a single extraction can generate undesired results, as, for example, a module that does not contain all axioms needed, or one that contains too many of them. Hence it is interesting to discover the relations between the extracted module with its superand sub-modules, because a slightly different module could be more interesting than the one extracted. In order to look at these (logic-based) relationships, a succint representation, called Atomic Decomposition, of the modular structure of an ontology has been introduced in [5]. In our paper [3], accepted for the Research Track at ISWC2011, we present the first, to our knowledge, large-scale investigation into decomposability and modular aspects of the NCBO BioPortal ontologies. We consider the Atomic Decompositions of such ontologies and demonstrate that most of them split into small logically coherent parts (atoms), from which modules can be efficiently assembled during reasoning. We discuss such good (on average) decomposability of BioPortal ontologies, its implications for applications, and also comment on