BERT and SVM Integration for Fake News Detection in Turkish: Evaluation with a New Dataset T rk e Sahte Haber Tespiti i in BERT ve SVM Entegrasyonu: Yeni Bir Veri K mesi ile De?gerlendirme

Cural N. M., Karabulut C., KARAGÖZ P.

33rd IEEE Conference on Signal Processing and Communications Applications, SIU 2025, İstanbul, Turkey, 25 - 28 June 2025, (Full Text) identifier

  • Publication Type: Conference Paper / Full Text
  • Doi Number: 10.1109/siu66497.2025.11112123
  • City: İstanbul
  • Country: Turkey
  • Keywords: BERTurk, convolutional neural networks, fake news detection, support vector machines, text classification, Turkish fact-checking
  • Middle East Technical University Affiliated: Yes

Abstract

This study presents an expanded new dataset and a hybrid BERT-SVM model for fake news detection in Turkish. As part of the research, news articles collected from FCTR, SOSYALAN, X-Fact datasets and Turkish fact-checking platforms teyit.org, and dogrulukpayi.com were combined to create a comprehensive dataset containing more than 20 thousand news articles. To improve classification accuracy, a hybrid approach integrating a fine-tuned BERTurk model with Support Vector Machines (SVM) was proposed. Additionally, model predictions were evaluated in terms of stylistic bias. The results demonstrate that BERT-based hybrid models have significant potential in addressing the unique challenges of fake news detection in Turkish.