Minimal Definition Signatures: Computation and Application to Ontology Alignment



Geleta, D
(2018) Minimal Definition Signatures: Computation and Application to Ontology Alignment. PhD thesis, University of Liverpool.

[img] Text
200674175_Jun2018.pdf - Unspecified

Download (2MB)

Abstract

In computer science, ontologies define a domain to facilitate knowledge representation and sharing, in a machine processable way. Ontologies approximate an actual world representation, and thus ontologies will differ for many reasons. Therefore knowledge sharing, and in general semantic interoperability, is inherently hindered or even precluded between heterogenous ontologies. Ontology matching addresses this fundamental issue by producing alignments, i.e. sets of correspondences that describe relations between semantically related entities of different ontologies. However, alignments are typically incomplete. In order to support and improve ontology alignment, and semantic interoperability in general, this thesis exploits the notion of implicit definability. Implicit definability is a semantic property of ontologies, signatures, and concepts (and roles) stating that whenever the signature is fixed under a given ontology then the definition of a particular concept (or role) is also fixed. This thesis introduces the notion of minimal definition signature (MDS) from which a given entity is implicitly definable, and presents a novel approach that provides an efficient way to compute in practice all MDSs of the definable entities. Furthermore, it investigates the application of MDSs in the context of alignment generation, evaluation, and negotiation (whereby agents cooperatively establish a mutually acceptable alignment to support opportunistic communication within open environments). As implicit definability permits defined entities to be removed without semantic loss, this thesis argues, that if the meaning of the defined entity is wholly fixed by the terms of its definition, only the terms in the definition are required to be mapped in order to map the defined entity itself; thus implicit definability entails a new type of definability-based correspondence correspondence. Therefore this thesis defines and explores the properties of definability- based correspondences, and extends several ontology alignment evaluation metrics in order to accommodate their assessment. As task signature coverage is a prerequisite of many knowledge-based tasks (e.g. service invocation), a definability-based, efficient approximation approach to obtaining minimal signature cover sets is presented. Moreover, this thesis outlines a specific alignment negotiation approach and shows that by considering definability, agents are better equipped to: (i) determine whether an alignment provides the necessary coverage to achieve a particular task (align the whole ontology, formulate a message or query); (ii) adhere to privacy and confidentiality constraints; and (iii) minimalise the cardinality of the resulting mutual alignment.

Item Type: Thesis (PhD)
Divisions: Fac of Science & Engineering > School of Electrical Engineering, Electronics and Computer Science
Depositing User: Symplectic Admin
Date Deposited: 15 Aug 2018 07:03
Last Modified: 03 Mar 2021 10:01
DOI: 10.17638/03022807
Supervisors:
URI: https://livrepository.liverpool.ac.uk/id/eprint/3022807