Modified and improved syntactic Structural and machine learning Approaches for ontology similarity Matching

Abstract

In semantic web, ontology models are reusable, organisations and individuals can more easily publish their own versions of ontologies. As a result, similar type of data can be represented in various ontologies resulting in data overlap, which could cause interoperability issues amongst web applications. Furthermore, the distributed nature of the internet causes data to be heterogeneous, posing a heterogeneity problem. An ontology matching solution can be a solution to alleviate these problems by discovering semantically related entities from two different ontologies. newlineIn the literature, several ontology matching approaches uses linguistic and syntactic matching, in which ontologies information such as ids, names, labels, descriptions, comments, and annotations are exploited. A few popular matching approaches found in the literature are AML, LogMap, Lily and Wiktionary Matcher. The gaps in the popular conventional approaches are as follows: newline(i) Conventional approaches are inefficient in synonym and hyponym based comparison. newline(ii) Partitioning ontologies may lead to lose certain semantic information. newline(iii) Selecting multiple external knowledge sources may create synonym conflicts between concepts. newline(iv) When training a huge dataset in a machine learning technique, annotating and labelling them becomes more critical. newline(v) The user intervention is required to validate the final alignments. newline