Detecting Communal Hate Speech in Hindi-English Code-Mixed Data on Social Media Using Machine Learning Models
International Journal of Drug Delivery Technology · 05.05.2026 · 16 (24s)
DOI: 10.25258/ijddt.16.24s.103 ↗Kurz gefasst
Fachartikel in der Zeitschrift International Journal of Drug Delivery Technology, online erschienen am 05.05.2026. Verfasst von Amit Sharma und 1 weiteren Autor:innen.
Abstract (Originalsprache)
The focus of the research is to investigate the identification of hate speech for code-mixed data that contains both Hindi and English. It is primarily concerned with political, religious, and social hostility on various social media platforms. We used a total of 10 datasets. There was a total of eight of them that were accessible to the public, two of which we collected manually from Instagram (4,322 comments) and YouTube (1,295 comments). In order to guarantee that the monitoring was fair and of a high standard, the comments were meticulously separated into two categories: 0 for non-hate speech that did not contain hate, and 1 for speech that contained hate. During the inquiry, there were two different experimental phases. For the purpose of constructing both traditional machine learning models (SVM, Logistic Regression, Naïve Bayes, KNN, and Decision Tree) and deep learning models (CNN and LSTM), we employed the technique of TF-IDF feature extraction methodology. The second phase consisted of the implementation of a Dual Contrastive Learning (DCL) framework. This framework was designed to solve the inadequacies of previous models in terms of identifying implicit hatred and context-dependent hostility in code-mixed information. The findings of the experiments demonstrated that DCL performed significantly better than both the ML and DL baselines, with average F1-scores ranging from 85% to 88%. This improvement shows that the model can detect small signs of community bias and language confusion, which are common in online spaces with many languages. Overall, the results show that models using contrastive learning work better at finding hate speech in complex, real-world social media data.
Zitieren
APA
Sharma, A., & Bhalla, R. (2026). Detecting Communal Hate Speech in Hindi-English Code-Mixed Data on Social Media Using Machine Learning Models. International Journal of Drug Delivery Technology, 16(24s). https://doi.org/10.25258/ijddt.16.24s.103
Vancouver
Sharma A, Bhalla R. Detecting Communal Hate Speech in Hindi-English Code-Mixed Data on Social Media Using Machine Learning Models. International Journal of Drug Delivery Technology. 2026;16(24s). doi:10.25258/ijddt.16.24s.103
BibTeX
@article{Sharma2026,
author = {Sharma, Amit and Bhalla, Rajni},
title = {Detecting Communal Hate Speech in Hindi-English Code-Mixed Data on Social Media Using Machine Learning Models},
journal = {International Journal of Drug Delivery Technology},
year = {2026},
volume = {16},
number = {24s},
doi = {10.25258/ijddt.16.24s.103}
}
EUVIMED