Use of ontologies for metadata records analysis in big data

Abstract

Big Data deals with the sets of information (structured, unstructured, or semi structured) so large that traditional ways and approaches (based on business intelligence decisions and database management systems) cannot be applied to them. Big Data is characterized by phenomenal acceleration of data accumulation and its complication. In different contexts Big Data often means both data of large volume and a set of tools and methods for their processing. Big Data sets are accompanied by metadata which contains a large amount of information about the data, including significant descriptive text information whose understanding by machines lead to better results of Big Data processing. Methods of artificial intelligence and intelligent Web-technologies improve the efficiency of all stages of Big Data processing. Most often this integration concerns the use of machine learning that provides the knowledge acquisition from Big Data and ontological analysis that formalizes for domain knowledge for Big Data analysis. In the paper, the authors present a method for analyzing the Big Data metadata which allows selecting those blocks of information among the heterogeneous sources and data repositories that are pertinent for solving the customer task. Much attention is paid to the matching of the text part of the metadata (metadata annotations) with the text describing the task. We suggest to use for these purposes the methods and instruments of natural language analysis and the Big Data ontology which contains knowledge about the specifics of this domain.