Multilevel ontologies for big data analysis and processing

Abstract

The problem of ever-increasing amounts of unstructured information in various fields of human activity is known as the problem of Big Data. Providing support for analytical activities requires determining the main factors that affect certain states of objects and processes in domains, as well as the degree of their influence, this significantly complicates the decision-making process, especially if data are represented heterogeneous information, there is a need to simultaneously take into account the impact of data from several areas dealing with several levels of classification. Given the significant volumes of text documents, it is impossible to solve the problem of structuring linguistic information by computer-aided extraction of the basic concepts that determine the text content (meaning), as well as the problem of constructing a formalized structure for formation the classes of individual objects and relations between them. The paper considers the ontological approach to the analysis and processing of Big Data represented both heterogeneous and linguistic data in the form of a multilevel ontology, implemented by computer-aided extracting of the basic concepts that define the text content (meaning) and determining semantic relations between the distributed information resources. The proposed approach uses the possibility of non-canonical conceptual ontologies to define equivalent concepts and thus to integrate the multiple ontologies that affect the same subject domain. This approach was implemented to create a multilevel ontology in the systemic biomedicine, the application of which in the process of postgraduate doctors and pharmacist’s education has significantly reduced the search time of relevant information and errors number due to the lack of unified terminology.