DETECTION OF SYNTACTIC ANDSEMANTIC REGULARITIES IN ONTOLOGIES

Abstract

Ontologies are machine processable artifacts and the core structures of theSemantic Web. OWL (Web Ontology Language) is a W3C Recommendation language fordeveloping ontologies; it is based on Description Logics, allowing for preciseknowledge representation and sound and complete automated reasoning over thecollection of axioms in an OWL document.Although ontologies are useful for sharing terminologies, their design and reuseare difficult and time consuming processes. Despite the efforts of the communitytowards the development of OWL ontologies, there is a lack of methods and toolsfor reusing and inspecting ontologies, i.e., reverse engineering methods.This thesis focuses on the area by investigating the detection ofregularities in ontologies, for the purpose of abstracting sets of axioms intopatterns that can be verified and reused. Its main contribution isthe Regularity Inspector for Ontologies (RIO) framework, whichimplements methods to find syntactic regularities (repetitive structuresin the asserted axioms) and semantic regularities (repetitivestructures in the entailments) in an ontology. Regularity detection is achieved through the use of cluster analysis for detectingsimilarities in sets of axioms. This thesis provides experimental evidence forthe effectiveness of regularity analysis for the inspection of patterns, and thediscovery of modeling irregularities (often modelling errors) during qualityassurance for real, large ontologies. In particular, empirical analysis showedthat RIO could successfully detect regularities in ontologies, revealing thepatterns adopted by the developers. It can be also used to trace patterndeviations as part of checking conformance to an intended design templateduring quality assurance of an ontology.This work has been motivated by the existence of pattern based systematicdevelopment methodologies and the lack of methods for discovering patterns inexisting ontologies---the natural complement of these pattern based developmentmethodologies.

Date of Award	1 Aug 2013
Original language	English
Awarding Institution	The University of Manchester
Supervisor	Robert Stevens (Supervisor) & Alan Rector (Supervisor)