Detecting fake news about Covid-19 using classifiers from Scikit-learn

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

Abstract

Weak control of reliability of news circulating on the Internet has provoked a large number of fake news and the need to develop detectors for such news. Unlike researchers that use deep learning networks with raw data, in this paper we consider popular machine learning algorithms from the Scikit-learn library and latent semantic indexes related to text to achieve good results in condition of a small dataset. The experiments use the full (10700) and small (1000) samples from the Constraint-2021 corpus, which includes news about Covid-19. With the best algorithms we achieved a detection quality of 78% and 71% micro F1-score, for the full and small datasets respectively. We believe that such a simple technology may be useful for users in their confrontation with fake news, because this approach does not demand a big dataset and can be implemented fast.

Original languageEnglish
Title of host publicationIEEE 16th International Conference on Computer Science and Information Technologies, CSIT 2021 - Proceedings
PublisherInstitute of Electrical and Electronics Engineers Inc.
Pages406-409
Number of pages4
ISBN (Electronic)9781665442572
DOIs
Publication statusPublished - 2021
Event16th IEEE International Conference on Computer Science and Information Technologies, CSIT 2021 - Lviv, Ukraine
Duration: 22 Sep 202125 Sep 2021

Publication series

NameInternational Scientific and Technical Conference on Computer Sciences and Information Technologies
Volume2
ISSN (Print)2766-3655
ISSN (Electronic)2766-3639

Conference

Conference16th IEEE International Conference on Computer Science and Information Technologies, CSIT 2021
Country/TerritoryUkraine
CityLviv
Period22/09/2125/09/21

Keywords

  • Classification
  • Constraint-2021
  • Fake news
  • Scikit-learn

Fingerprint

Dive into the research topics of 'Detecting fake news about Covid-19 using classifiers from Scikit-learn'. Together they form a unique fingerprint.

Cite this