Ontology matching in a distributed environment

Nowadays, society, research and industries are faced with a continuously growing amount of information available as structured or unstructured data presenting semantic meanings. In the scope of this work, data structures based on semantic specifications are investigated concluding in an approach for receiving customized semantic data related to a certain domain of interest. As ontologies, originally a philosophy discipline for identifying the fundamental structure of reality, are similar to interlinked semantic data collections extended with rules. They are used for storing, updating and utilizing semantic data. Such ontologies have already reached the barrier of billions of interlinked concepts, causing difficulties in terms of performance when trying to execute queries. Computer sciences define an ontology as a conceptualization of formal representations including interconnected concepts related to a certain domain. Through the matching of several ontologies relevant semantic data containing information are identified or the validity of data structures is proven by matching it to each other. Current ontology matching approaches are based on various matching strategies being the base for several tools and frameworks. Hence, this work is based on such strategies as well as latest ontology matching tools and frameworks. This work investigated an enhanced ontology matching approach dealing with the continuously growing amount of semantic data for generating facts related to a certain domain of interest. Thus, the processing of semantic data face the challenge of allocating sufficient computing resources as well as for performing a content wise sensible matching of needed data. Current approaches struggle with performance and efficiency issues as soon as the semantic data reaches a high amount of data, e.g. in August 2011 in the scope of the large triple store challenge [1], the loading and querying of about one trillion Resource Description Framework (RDF) triples [2] was processed. The execution took nearly 338 hours by an average rate of 829.556 RDF triples per second running on 80 cores. In addition, the generation of precise matching results needs to be considered. Current risks in the scope of matching semantic data are the generation of false matching results due to an insufficient matching, not accurate enough to receive high quality results, and an afterwards performed reasoning based on the matching results not suitable when thinking of big interlinked data volumes. Additionally, a reasoning based on a corrupt data set generated by an insufficient matching will produce wrong facts. Thus, the main question of this work is how to bridge the gap between current matching approaches and large semantic data stores by applying a semantic matching usable for reasoning on those big data stores. Such a proceeding will manage the continuously growing immense amount of data through a matching strategy and thus, providing an individually customized