Optimizing Social Bot Detection Via Transformer-Oriented Classification

Read full text Summarise with VIRAT
Authors: Mrs. Moorpa Rekha, Mr. Bommineni Srinivasulu, Arkaru Vedavathi, Bhogadi Supriya, Putturu Uday Kiran, Kalivillolu Yaswanth
Cited by: 0

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

The rapid proliferation of online communication platforms and social media has led to the emergence of malicious social bots that mimic human behavior, disseminate misinformation, and compromise the integrity of digital interactions. Distinguishing human-generated text from machine-generated content has become increasingly difficult with the advancement of deep neural networks. Transformer-based Pre-trained Language Models (PLMs) have recently demonstrated remarkable performance in Natural Language Understanding (NLU), making them a promising foundation for bot detection. This study proposes a tweet-level bot detection framework that leverages fine-tuned PLMs, such as BERT, RoBERTa, and GPT-3, to generate high-quality contextual embeddings. A Feedforward Neural Network (FNN) is applied on top of these embeddings for final classification. Experimental evaluation on the Twitter bot dataset achieved an F1-score of 95%, surpassing traditional methods including Word2Vec and GloVe. Furthermore, Explainable Artificial Intelligence (XAI) was incorporated to enhance interpretability, reliability, and trust in the classification outcomes.To improve robustness, scalability, and adaptability, the proposed work introduces several extensions: (i) multi-modal detection using textual, profile, and network features; (ii) ensemble learning with hybrid transformers; (iii) adversarial training for robust defense; (iv) real-time detection for streaming data; (v) cross-platform generalization; and (vi) temporal behavior modeling. These enhancements enable a more resilient and transparent approach, offering a comprehensive solution to mitigate the growing threat of social bots across diverse online ecosystems.

Fields of Study & Subjects

Spam and Phishing Detection Misinformation and Its Impacts Hate Speech and Cyberbullying Detection

Publication details

Journal / source
Zenodo (CERN European Organization for Nuclear Research)
Publication date
22 March 2026
Language
EN