Expert, Journal, and Automatic Classification of Full Texts and Annotations of Scientific Articles
Irina Selivanova, Denis V. Kosyakov, D. A. Dubovitskii, Andrey E. Guskov · Automatic Documentation and Mathematical Linguistics · 2021
In this article we consider a fundamentally new information-theoretic approach to the classification of scientific texts based on compression algorithms. An analysis using the example of the comparative classification of full-text documents from arXiv.org and short annotations from Scopus showed that the accuracy of the proposed method is 87–92% and, in general, is not inferior to the existing ones. These conclusions were confirmed by an expert assessment.