Methods and Tools for Virtual Semantic Integration of Data from Distributed Heterogeneous Sources

Abstract

The article is devoted to the natural language processing from distributed heterogeneous sources based on the principles of their virtual semantic integration. The main purpose of data integration is to provide the user with unified access to distributed data as a single virtual storage for performing natural language queries, regardless of the data storage format and location. The article discusses the main approaches focused on virtual semantic data integration, and describes the proposed concept of building an ontology driven instrumental environment based on Data Fabric technology, which allows to automate data processing via intermediate layer of ontologies in a unified form. The article describes NuCoBoShell that is the instrumental environment implementing the proposed approach. NuCoBoShell uses ontology-driven semantic integration mechanism to provide the answering, which, unlike traditional Internet answering services, provides the opportunity to obtain more pertinent answers automatically extracting the necessary information from not only heterogeneous web resources, but also text documents stored in accessible data warehouses and user's local computer without the need to copy data to a single repository.