Survey on Hate Speech Detection in Hindi English Code Mixed Language: Datasets, Methods and Future Directions

Online social media applications have encouraged users to exercise their freedom of expression. However, this freedom also brings responsibility. Some users misuse these platforms to spread hatred and hostility. Hate speech typically targets groups or individuals based on religion, community, gender, caste, race, or other discriminatory attributes. In Hinglish NLP, datasets and shared tasks often use partially overlapping labels such as hate, offensive/abusive, hostility, aggression, and toxicity. To avoid conflating these tasks, we adopt an operational taxonomy that separates (i) hate speech (protected‐group targeted), (ii) offensive/abusive language (non‐protected insults/profanity), and (iii) aggression/toxicity (harmful intent/uncivil language that may be non‐targeted). We map each reviewed dataset and paper to this taxonomy and treat aggression/toxicity corpora as related but distinct resources rather than hate‐speech corpora. While most prior research has predominantly focused on the English language for hate speech detection, this survey focuses on the code‐mixed language in Hindi‐English, commonly referred to as Hinglish. In this survey paper, we outline our methodology for identifying relevant prior works. Given that hate speech detection is largely driven by data‐centric computational methods, we analyze trends in the curated datasets, moving from simple binary (hate/non‐hate) to more complex label targets, types, and severity of hate. We study preprocessing techniques specifically designed for handling Hindi‐English code‐mixed text. Next, we present trends in the computational pipeline moving from traditional‐based machine learning approaches toward deep learning and transformer‐based approaches to hate speech detection in Hinglish. Despite such advancements, future research needs to address some of the significant issues, such as model generalization, temporal dynamics of hate speech, severity classification, analysis of counter‐speech, and ensuring fairness in hate speech detection models for Hinglish code‐mixed text.

Authors

Institutions

Publication Details

Journal
Wiley Interdisciplinary Reviews Data Mining and Knowledge Discovery
Published
2026-11-20
DOI
https://doi.org/10.1002/widm.70142
Primary Topic
Hate Speech and Cyberbullying Detection
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Survey on Hate Speech Detection in Hindi English Code Mixed Language: Datasets, Methods and Future Directions

Rishabh Kaushal, Kalpana Yadav, Jyoti Sharma
Wiley Interdisciplinary Reviews Data Mining and Knowledge Discovery
Hate Speech and Cyberbullying Detection
article

Survey on Hate Speech Detection in Hindi English Code Mixed Language: Datasets, Methods and Future Directions

Rishabh Kaushal, Kalpana Yadav, Jyoti Sharma
article en

Abstract

Online social media applications have encouraged users to exercise their freedom of expression. However, this freedom also brings responsibility. Some users misuse these platforms to spread hatred and hostility. Hate speech typically targets groups or individuals based on religion, community, gender, caste, race, or other discriminatory attributes. In Hinglish NLP, datasets and shared tasks often use partially overlapping labels such as hate, offensive/abusive, hostility, aggression, and toxicity. To avoid conflating these tasks, we adopt an operational taxonomy that separates (i) hate speech (protected‐group targeted), (ii) offensive/abusive language (non‐protected insults/profanity), and (iii) aggression/toxicity (harmful intent/uncivil language that may be non‐targeted). We map each reviewed dataset and paper to this taxonomy and treat aggression/toxicity corpora as related but distinct resources rather than hate‐speech corpora. While most prior research has predominantly focused on the English language for hate speech detection, this survey focuses on the code‐mixed language in Hindi‐English, commonly referred to as Hinglish. In this survey paper, we outline our methodology for identifying relevant prior works. Given that hate speech detection is largely driven by data‐centric computational methods, we analyze trends in the curated datasets, moving from simple binary (hate/non‐hate) to more complex label targets, types, and severity of hate. We study preprocessing techniques specifically designed for handling Hindi‐English code‐mixed text. Next, we present trends in the computational pipeline moving from traditional‐based machine learning approaches toward deep learning and transformer‐based approaches to hate speech detection in Hinglish. Despite such advancements, future research needs to address some of the significant issues, such as model generalization, temporal dynamics of hate speech, severity classification, analysis of counter‐speech, and ensuring fairness in hate speech detection models for Hinglish code‐mixed text.

Wiley Interdisciplinary Reviews Data Mining and Knowledge DiscoveryVol. 16(4)
Indira Gandhi Delhi Technical University for Women (IN)
Hate Speech and Cyberbullying Detection
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon