Informatics and Applications

2022, Volume 16, Issue 4, pp 63-72

TECHNOLOGY FOR CLASSIFICATION OF CONTENT TYPES OF E-TEXTBOOKS

  • A. V. Bosov
  • A. V. Ivanov

Abstract

The problem of automatic classification of the educational content of the e-learning system, represented by tasks or practical examples, is being solved. A promising direction in the development of e-learning systems is the assessment of the quality of educational content. Carrying out such an assessment is the rationale for the need to create an automated classifier. The main idea is to model the content with an object with two properties - a textual description in natural language and a set of formulas in the language of scientific computer layout TgX. Using tasks from the electronic textbook on the theory of functions of a complex variable, a data set was prepared and labeled in accordance with this model. Four text classification algorithms were trained - naive Bayes classifier, logistic regression, single-layer and multilayer feedforward neural networks. For these classifiers, a number of comparative experiments were carried out comparing the classification accuracy using text content only, formula content only, and the full model. As a result of the experiment, not only a formal comparison of the algorithms was carried out but also the fundamental advantage of the full model was shown. That is, when using both textual description and representation of formulas in the TjXlanguage, the classification accuracy significantly exceeds one-factor algorithms and confirms the readiness of the technology for practical application.

[+] References (36)

[+] About this article