Work

LLM ‐Assisted Topic Modelling for Hate Speech Characterization

Year: 2026

Type: article

Abstract: ABSTRACT In the digital era, the internet and social media have transformed communication but have also facilitated the spread of hate speech and disinformation, leading to radicalization, polarisation and toxicity. This is especially concerning for media outlets due to their significant role in shaping public discourse. This study examines the topics, sentiments, and hate prevalence in response messages (website comments and tweets) to news from five Spanish media outlets (La Vanguardia, ABC, El País, El Mundo and 20 Minutos) in January 2021. These public reactions were originally labelled as distinct types of hate by experts following an original procedure, and they are now classified into three sentiment values (negative, neutral, or positive) and main topics. The BERTopic unsupervised framework was used to extract topics, manually named with the help of large language models (LLMs) and grouped into nine primary categories. Results show social issues (), expressions and slang (), and political issues () as the most discussed. Content is mainly negative () and neutral (), with low positivity (). Toxic narratives relate to conversation expressions, gender, feminism and COVID‐19. Despite low levels of hate speech (), the study confirms high toxicity in online responses to social and political topics.

Source: Expert Systems

DOI: 10.1111/exsy.70293

Authors: Alejandro Buitrago López, Javier Pastor‐Galindo, José A. Ruipérez‐Valiente

Institution: Universidad de Murcia

Language: English

Topic: Hate Speech and Cyberbullying Detection

Subfield: Artificial Intelligence

Field: Computer Science

Domain: Physical Sciences

Open Access status: hybrid

Estimated APC paid: $3,600

Funders: Universidad de Murcia, Instituto Nacional de Ciberseguridad