Examinando por Autor "Beres, Sean Robert"

Mostrando 1 - 1 de 1
  • Miniatura
    Publicación
    Social Bot Detection Using Deep Learning and Linguistic Features: A State-of-the-Art Literature Review
    (Universidad Nacional de Educación a Distancia (UNED). Facultad de Filología, 2025-03-11) Beres, Sean Robert; Escobar Álvarez, Mª Ángeles
    In the current online landscape, social media bots have evolved from simple, rule-based software to sophisticated LLM-assisted entities capable of mimicking human interaction. Detection methods relying on metadata-based features have become increasingly ineffective against these new bot models, necessitating an exploration of linguistic and content-based approaches. This work serves as a primer on the social bot detection task for linguists as well as a semi-systematic qualitative literature review on state-of-the-art social bot detection. The former includes a discussion of social impact, different types of deep learning models, feature sets, and linguistic indicators of synthetic text. By examining 14 examples of cutting-edge research published from 2022 to 2024, the literature review assesses the efficacy of neural network architecture leveraging linguistic features in identifying social bot activity. The review highlights key challenges, such as adversarial evasion tactics, computational overhead, and ethical considerations surrounding privacy and false positives. Toward this end, datasets, success rates in F1-score, model architectures, and handcrafted features have been extracted and explored. Furthermore, the study discusses the role that training data, linguistic embeddings, and learning styles play in improving detection models. This work’s main findings include the need for dataset coherency, scalable, real-time frameworks, unsupervised or self-supervised learning styles, and model explainability.