Deep Learning Approach for Author Verification Problem on Twitter

Yilmaz M., MUTLU B., Utku A., AKCAYOL M. A.

26th IEEE Signal Processing and Communications Applications Conference (SIU), İzmir, Türkiye, 2 - 05 Mayıs 2018 identifier identifier

  • Yayın Türü: Bildiri / Tam Metin Bildiri
  • Cilt numarası:
  • Doi Numarası: 10.1109/siu.2018.8404365
  • Basıldığı Şehir: İzmir
  • Basıldığı Ülke: Türkiye
  • Gazi Üniversitesi Adresli: Evet

Özet

In social media, there are a large number of parody accounts belong to well known people. It is highly important to distinguish a sharing posted by these kind of accounts. In literature, this problem, referred to as author verification, has been investigated by basing traditional machine learning methods. However, the behavior of deep learning methods on social media sharing, which is composed of limited text length, has not been studied to solve this problem. In this study, a deep learning approach based on multilayer perceptron has been presented for author verification problem on Twitter data. As a result of experiments, high classification success (accuracy: 92.083%, precision: 77.894%, recall: 88.58%, ROC-AUC: 95.95%) has been obtained.