Improved algorithms for module extraction and atomic decomposition

In recent years modules have frequently been used for ontology development and understanding. This happens because a module captures all the knowledge an ontology contains in a given area, and often is much smaller than the whole ontology. One useful modularisation technique for expressive ontology languages is locality-based modularisation, which allows for fast (polynomial) extraction of modules. In order to better understand the modular structure of an ontology, a technique called Atomic Decomposition can be used. It efficiently builds a structure representing all possible modules for an ontology, regardless of the modularisation algorithm adopted and without the need to compute an exponential number of modules, as in a naive approach. This structure may be used e.g., for quick extraction of modules, or to investigate dependencies between modules, and so on. However, existing algorithms for both locality-based module extraction and atomic decomposition do not scale well. This happens mainly because of their global nature: each iteration always explores the whole ontology, even when it is not necessary. We propose algorithms for locality-based module extraction and atomic decomposition that work only on the relevant part of the ontology. This improves performance of algorithms by avoiding unnecessary checks. Empirical evaluation confirms a significant speed up on real-life ontologies.