Persian ontology matching: Challenges, dataset generation and similarity combination

Abstract

This paper presents a practical study of the Persian ontology matching. Ontology matching has a key role to develop the semantic web. Although many attempts are done to develop Persian ontologies, but the Persian ontology matching problem is still unresolved. This paper addresses the challenges of the Persian ontology matching. One of the most important prerequisites of design and develop efficient ontology matchers is standard benchmark datasets that allow a fair evaluation and comparison between different matchers. First, we generated a benchmark dataset for Persian ontology matching that we named it PersianFarm. PersianFarm is developed according to OntoFarm, the multilingual dataset of the Ontology Alignment Evaluation Initiative (OAEI). It consists of seven Persian ontologies and eleven reference alignments between them. Next, we evaluate a wide range of similarity metrics such as string based, structural and context-based similarities against PersianFarm dataset. Finally, different similarity metrics have been selected and combined to develop an appropriate Persian ontology matcher. The results that reported as F-measure rate, show that the mixture of similarities achieved reasonable results to match the concepts.