Ontological Approach: Knowledge Representation and Knowledge Extraction
- PDF / 622,773 Bytes
- 11 Pages / 612 x 792 pts (letter) Page_size
- 32 Downloads / 204 Views
Ontological Approach: Knowledge Representation and Knowledge Extraction O. M. Ataeva1* , V. A. Serebryakov1** , and N. P. Tuchkova1*** (Submitted by A. M. Elizarov) 1
Federal Research Center ”Computer Sciences and Control,” Russian Academy of Sciences, Moscow, 119333 Russia Received April 4, 2020; revised April 15, 2020; accepted April 19, 2020
Abstract—The application of artificial intelligence algorithms for data analysis, characteristics, and metrics of scientific information resources are considered. In this paper, we discuss how metrics are related to assessment of scientific publication components and whether metrics are related to fundamental knowledge. It was noted that the characteristics of professional scientific activity are evaluated on the basis of metrics that are not related to the knowledge characteristics. The problem of knowledge extraction was studied on the basis of data verification by means of logical evidence– based schemes specified in the knowledge ontology. Properties of the modern stage of development of the knowledge space as a resource of artificial intelligence were noted. The transformation of artificial intelligence tasks into a new digital age was also analyzed. The insufficient use of artificial intelligence and machine learning methods in scientific bibliographic databases was emphasized, where quantitative scientometric indicators prevailed. Examples of ontological presentation of data and knowledge extraction are discussed and the special role of ontological approach to data structuring and knowledge extraction is highlighted. DOI: 10.1134/S1995080220100030 Keywords and phrases: knowledge extraction, knowledge base, knowledge space, ontology of subject domain, artificial intelligence methods, knowledge metrics.
1. INTRODUCTION The challenges of modern information technology (IT) are associated with the need to implement methods of artificial intelligence (AI) in the systems of scientific knowledge. Common methods of analyzing arrays of scientific publications relate more to the assessment of the publication activity level of authors. To evaluate the significance of the publication, scientometric (quantitative) indicators are proposed to be used in the form of statistics on citation, H-index, impact factor of publication, quartile of journal (Q1–Q4) [1–7], and others. These estimates are typical for the period of accumulation of information resources. Nevertheless, ignoring the real values of knowledge accumulated in the databases of publications means lowering the scientific significance of the studies reflected in these works. The problem is of extracting useful and significant knowledge from these resources with regards to science and expanding the corresponding subject ontologies through new scientific facts and links. Knowledge, as structured data [8], is precisely stated in scientific publications. It is necessary to evaluate how much they will serve to enrich the subject area and whether the results presented in the publications are new, copyrighted, reliable, etc. There is a whole range of is
Data Loading...
