Ontology management and selection in re-use scenarios

• Main Author: Finney, KT
• Format: Thesis
• Language: English
• Published: 2012
• Subjects: ontology; management; evaluation; feature type; feature catalogue; Antarctic; Crabtree; Antarc*
• Online Access: https://eprints.utas.edu.au/15912/; https://eprints.utas.edu.au/15912/1/whole-finny-2012.pdf; https://eprints.utas.edu.au/15912/2/front-finny-2012.pdf; https://eprints.utas.edu.au/15912/3/OwlFiles.zip

One of the main impediments to realising the Semantic Web vision is that most scientific data, even those data deployed on the web, are not generally expressed or encoded in an unambiguously defined, machine-interpretable manner. This is particularly the case for Antarctic-themed data. Ontologies that are linked to datasets via semantic annotation are required to achieve semantic-enablement of scientific data infrastructure. In scientific communities that adhere to the Open Geospatial Consortium Service-Oriented-Architecture (Web services) paradigm, Feature Catalogues are the repositories intended to manage and publish descriptions of dataset concepts. This thesis explores how Feature Catalogues can be ontologically-grounded to facilitate semantic annotation and in doing so addresses the lack of guidance in current standards about how to configure an ontologically grounded Feature Catalogue and how best to access the resources it contains for the semantic annotation of Web services. Also investigated is how ontology selection and evaluation is currently taking place in practise because ontology evaluation methodologies mentioned in the literature are resource intensive to apply, often requiring a high level of ontological expertise. Both contributions seek to lower barriers for ontology uptake and reuse within scientific communities. To address these issues, two scientific communities of practise (i.e., AODN and SCAR) were used as case studies within a Design Science research method to ground-truth the design and to prototype an ontologically grounded, service-enabled Feature Catalogue. To address research questions pertaining to ontology selection and evaluation practise, fourteen experts (from outside of the AODN and SCAR communities) with experience in building semantically-enabled scientific infrastructure, were surveyed and interviewed to ascertain what ontology evaluation methods and criteria are being used in practise. A hierarchical evaluation model was established from analysed expert data using Template Analysis (Crabtree and Miller, 1992; King, 2004). The Analytical Hierarchical Processing (AHP) technique (Saaty, 1980), was then harnessed to establish the relative importance given by experts to each of the model elements. The contributions arising consisted of an enhanced ISO 19110 Feature Catalogue model which accommodated additional concepts necessary to describe the observation-centric dataset paradigms of the two case study communities. The extended conceptual model was semantically grounded using the DOLCE (upper ontology) and expressed in both OWL and SKOS. Demonstration REST-based service interfaces (and REST query patterns) were created for serving Catalogue content to requesting Web clients. To the author’s knowledge, no other Feature Catalogue implementation, founded on the ISO 19110 conceptual model, has attempted to model the Catalogue as an ontology, or permits access to Catalogue content via REST-based service interfaces. This thesis also delivers a “practical” framework for evaluating and then selecting reusable ontological content which encompasses weighted model elements (indicating relative levels of importance), coupled with expert-derived evaluation metrics. Although the evaluation criteria listed in the framework are not novel in themselves, identifying which criteria are of most utility to experts who are operating in real-world scenarios, is an important contribution to practise.