Enhanced systems for ontology matching evolution and learning from text

Abstract

The quantum of accessible heterogeneous textual information has newlinegreatly proliferated, owing to increased web use. This calls for a newlinerepresentation that semantically consolidates and organizes information in a newlineconceptual hierarchy so as to store, retrieve and infer knowledge from a range newlineof sources. Ontology is the best candidate for this sort of representation. newlineOntologies are acquired using a range of building methods such as ontology newlinematching, ontology evolution, and ontology learning from text. Although newlinenumerous ontology building systems exist in the literature, there still are newlinenumerous open challenges to be tackled, such as the efficient matching of newlineontologies, automated ontology evolution, and handling knowledge newlineacquisition bottlenecks for constructing quality ontologies. This calls for newlineimprovements in the performance of the building methods. In this thesis, newlinevarious measures and methods are designed to improve the performance of newlinethe ontology matching, evolving, and learning from text systems. newlineOntology matching is an effective way to enable interoperability newlineamong the numerous and potentially complementing or conflicting ontologies newlineof same domain. Efficiency of ontology matching is an important criteria newlinesince matching is a time and resource consuming process. newlineIn this thesis, efficiency is improved by designing a partitioningbased newlinematching system where ontologies are partitioned into sub-ontologies newlineusing a novel static neighbour-based similarity measure and a partitioning newlinealgorithm. Here, only similar sub-ontology pairs across two newline newline