A Large Language Model-Based Approach for Multilingual Hate Speech Detection on Social Media

The proliferation of hate speech on social media platforms poses significant threats to digital safety, social cohesion, and freedom of expression. Detecting such content—especially across diverse languages—remains a challenging task due to linguistic complexity, cultural context, and resource limit...

Full description

Saved in:
Bibliographic Details
Main Authors: Muhammad Usman, Muhammad Ahmad, Grigori Sidorov, Irina Gelbukh, Rolando Quintero Tellez
Format: Article
Language:English
Published: MDPI AG 2025-07-01
Series:Computers
Subjects:
Online Access:https://www.mdpi.com/2073-431X/14/7/279
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1849406601360834560
author Muhammad Usman
Muhammad Ahmad
Grigori Sidorov
Irina Gelbukh
Rolando Quintero Tellez
author_facet Muhammad Usman
Muhammad Ahmad
Grigori Sidorov
Irina Gelbukh
Rolando Quintero Tellez
author_sort Muhammad Usman
collection DOAJ
description The proliferation of hate speech on social media platforms poses significant threats to digital safety, social cohesion, and freedom of expression. Detecting such content—especially across diverse languages—remains a challenging task due to linguistic complexity, cultural context, and resource limitations. To address these challenges, this study introduces a comprehensive approach for multilingual hate speech detection. To facilitate robust hate speech detection across diverse languages, this study makes several key contributions. First, we created a novel trilingual hate speech dataset consisting of 10,193 manually annotated tweets in English, Spanish, and Urdu. Second, we applied two innovative techniques—joint multilingual and translation-based approaches—for cross-lingual hate speech detection that have not been previously explored for these languages. Third, we developed detailed hate speech annotation guidelines tailored specifically to all three languages to ensure consistent and high-quality labeling. Finally, we conducted 41 experiments employing machine learning models with TF–IDF features, deep learning models utilizing FastText and GloVe embeddings, and transformer-based models leveraging advanced contextual embeddings to comprehensively evaluate our approach. Additionally, we employed a large language model with advanced contextual embeddings to identify the best solution for the hate speech detection task. The experimental results showed that our GPT-3.5-turbo model significantly outperforms strong baselines, achieving up to an 8% improvement over XLM-R in Urdu hate speech detection and an average gain of 4% across all three languages. This research not only contributes a high-quality multilingual dataset but also offers a scalable and inclusive framework for hate speech detection in underrepresented languages.
format Article
id doaj-art-09e208cdd660403a8e1a904d5e0c84c9
institution Kabale University
issn 2073-431X
language English
publishDate 2025-07-01
publisher MDPI AG
record_format Article
series Computers
spelling doaj-art-09e208cdd660403a8e1a904d5e0c84c92025-08-20T03:36:19ZengMDPI AGComputers2073-431X2025-07-0114727910.3390/computers14070279A Large Language Model-Based Approach for Multilingual Hate Speech Detection on Social MediaMuhammad Usman0Muhammad Ahmad1Grigori Sidorov2Irina Gelbukh3Rolando Quintero Tellez4Centro de Investigación en Computación, Instituto Politécnico Nacional (CIC-IPN), Mexico City 07320, MexicoCentro de Investigación en Computación, Instituto Politécnico Nacional (CIC-IPN), Mexico City 07320, MexicoCentro de Investigación en Computación, Instituto Politécnico Nacional (CIC-IPN), Mexico City 07320, MexicoCentro de Investigación en Computación, Instituto Politécnico Nacional (CIC-IPN), Mexico City 07320, MexicoCentro de Investigación en Computación, Instituto Politécnico Nacional (CIC-IPN), Mexico City 07320, MexicoThe proliferation of hate speech on social media platforms poses significant threats to digital safety, social cohesion, and freedom of expression. Detecting such content—especially across diverse languages—remains a challenging task due to linguistic complexity, cultural context, and resource limitations. To address these challenges, this study introduces a comprehensive approach for multilingual hate speech detection. To facilitate robust hate speech detection across diverse languages, this study makes several key contributions. First, we created a novel trilingual hate speech dataset consisting of 10,193 manually annotated tweets in English, Spanish, and Urdu. Second, we applied two innovative techniques—joint multilingual and translation-based approaches—for cross-lingual hate speech detection that have not been previously explored for these languages. Third, we developed detailed hate speech annotation guidelines tailored specifically to all three languages to ensure consistent and high-quality labeling. Finally, we conducted 41 experiments employing machine learning models with TF–IDF features, deep learning models utilizing FastText and GloVe embeddings, and transformer-based models leveraging advanced contextual embeddings to comprehensively evaluate our approach. Additionally, we employed a large language model with advanced contextual embeddings to identify the best solution for the hate speech detection task. The experimental results showed that our GPT-3.5-turbo model significantly outperforms strong baselines, achieving up to an 8% improvement over XLM-R in Urdu hate speech detection and an average gain of 4% across all three languages. This research not only contributes a high-quality multilingual dataset but also offers a scalable and inclusive framework for hate speech detection in underrepresented languages.https://www.mdpi.com/2073-431X/14/7/279large language modelsGPTmultilingual hate speech detectiondata miningmachine learningdeep learning
spellingShingle Muhammad Usman
Muhammad Ahmad
Grigori Sidorov
Irina Gelbukh
Rolando Quintero Tellez
A Large Language Model-Based Approach for Multilingual Hate Speech Detection on Social Media
Computers
large language models
GPT
multilingual hate speech detection
data mining
machine learning
deep learning
title A Large Language Model-Based Approach for Multilingual Hate Speech Detection on Social Media
title_full A Large Language Model-Based Approach for Multilingual Hate Speech Detection on Social Media
title_fullStr A Large Language Model-Based Approach for Multilingual Hate Speech Detection on Social Media
title_full_unstemmed A Large Language Model-Based Approach for Multilingual Hate Speech Detection on Social Media
title_short A Large Language Model-Based Approach for Multilingual Hate Speech Detection on Social Media
title_sort large language model based approach for multilingual hate speech detection on social media
topic large language models
GPT
multilingual hate speech detection
data mining
machine learning
deep learning
url https://www.mdpi.com/2073-431X/14/7/279
work_keys_str_mv AT muhammadusman alargelanguagemodelbasedapproachformultilingualhatespeechdetectiononsocialmedia
AT muhammadahmad alargelanguagemodelbasedapproachformultilingualhatespeechdetectiononsocialmedia
AT grigorisidorov alargelanguagemodelbasedapproachformultilingualhatespeechdetectiononsocialmedia
AT irinagelbukh alargelanguagemodelbasedapproachformultilingualhatespeechdetectiononsocialmedia
AT rolandoquinterotellez alargelanguagemodelbasedapproachformultilingualhatespeechdetectiononsocialmedia
AT muhammadusman largelanguagemodelbasedapproachformultilingualhatespeechdetectiononsocialmedia
AT muhammadahmad largelanguagemodelbasedapproachformultilingualhatespeechdetectiononsocialmedia
AT grigorisidorov largelanguagemodelbasedapproachformultilingualhatespeechdetectiononsocialmedia
AT irinagelbukh largelanguagemodelbasedapproachformultilingualhatespeechdetectiononsocialmedia
AT rolandoquinterotellez largelanguagemodelbasedapproachformultilingualhatespeechdetectiononsocialmedia