Rough Set Based Ontology Matching

Introduction

The heterogeneity nature of data in semantic web has led to innovations in efficient representation and data retrieval. Knowledge representation play a major role in semantic web as it enhances information reuse. Ontologies support the task of capturing, processing and reuse of information in the form of individuals, classes, attributes, relations and instances within a domain. The growth of semantic web augments the development of more number of ontologies for a single domain. Ontological representation done by different designers in different perspectives paved way for the growth of ontology matching. Ontology matching is the process of finding the relations between semantically related entities of different ontologies. It aims in finding correspondences among entities present in the source ontology and target ontology considered. The significant components that are to be focused in the task of ontology matching include concepts, its corresponding attributes and the relationships among those concepts. The matching task focus on these components to determine matches by utilizing similarity matchers to obtain correspondences. The large size of ontologies which brings in the need for search space optimization in ontology matching. Several researches were done to handle search space reduction among which utilization of concept importance factor is one of the most trivial techniques. The concept importance metric identifies the significant concepts present in the ontologies that can be considered for matching. The semantically related entities are obtained by determining similarities using similarity matchers which are responsible for computing similarity at various levels ranging from string to the structural level. Single similarity matcher or multiple similarity matchers may be employed depending on the matching situation. The aim of ontology matching is to obtain maximum number of matches and hence aggregation of various matchers is utilized for achieving increased precision results. The aggregation of the similarity values obtained from similarity matches leads to a mystification in determination of matches leading to uncertainty. Uncertainty in ontology matching, is a condition where the result of matching can neither be termed as similar not dissimilar. It indicates the output obtained as a result of the ontology matching process cannot be termed as the most appropriate or accurate. Several ontology matching problems exist as discussed by Arthi and Saruladha (2015) but uncertainty is a less handled issue.

This paper illustrates the related works in Section 2 which briefs about the various techniques that has been proposed to handle ontology matching. Section 4 depicts the theoretical background and its application in handling uncertainty in ontology matching. Section 5 presents the basic concepts that are used in the RSOM system and Section 6 briefs the complete working of the RSOM system with its algorithm, the similarity matchers used along with the technique for handling uncertainty. Section 7 illustrates the application of the proposed RSOM system to perform ontology matching for a sample set of concepts in the OAEI benchmark dataset (Russia 1 and Russia 2). Section 8 describes the accuracy of the proposed system using the experimental results and Section 9 briefs on the performance metrics and two other proposed performance metrics used for evaluating the proposed system. Section 10 describes the conclusion and future enhancements are to be handled.