OntoElecting Requirements for Domain Ontologies

The Case of Time Domain

  • Vadim Ermolayev Department of Computer Science, Zaporizhzhia National University, Ukraine
Keywords: Requirements Elicitation, Automated Term Extraction from Text, Feature Conceptualization, Ontology Engineering, Ontology Refinement, Ontology Evaluation, Ontology Fitness, Domain Knowledge Stakeholder, Vote, OntoElect

Abstract

This paper reports on the use of the OntoElect methodology for evaluating the fitness of an existing ontology to the requirements of the knowledge stakeholders in a domain. It demonstrates, that a thorough routine for indirect elicitation, ensuring completeness, correctness of interpretation, using in ontology evaluation of these requirements is a must for ontology engineering. This is also valid if the requirements for ontology refinement are elaborated by a very high profile expert working groups. The approach used in the reported research is based on the use of OntoElect – the methodology for ontology refinement. The workflow of OntoElect contains three phases: feature elicitation, requirements conceptualization, and ontology evaluation. It elicits the set of terms extracted from a saturated collection of documents in the domain. It further sublimates these terms to the set of required features using the information about term significance in the form of numeric scores. Furthermore, it applies conceptualization and formalization activities to these features yielding their aggregations as ontological fragments interpreted as formalized requirements. Finally, the mappings are specified between the elements in the requirements and ontology elements. The scores are used in the mappings to indicate the strength of positive or negative votes regarding the evaluated ontology. The sum of the votes gives the overall numeric fitness measure of the ontology to the domain requirements. The paper presents the use of OntoElect in the use case of evaluating the W3C OWL-Time ontology against the requirements extracted from the proceedings of the TIME symposia series.

Published
2018-02-27
Section
Invited Contribution