Bengali dataset and transformer model detect anti male hate speech
Ben-Misandry: A Bengali Dataset and Late-Fusion Transformer Framework for Detecting Anti-Male Gender Hate
Computation and Language
Summary
Hate speech research in the Bengali language has mostly focused on attacks against women, ignoring hate against men. The authors created the first Bengali dataset specifically for detecting hate speech targeting men, including mocking male emotional sensitivity. They used AI to generate 5,000 example comments and tested various computer models to classify them. Their best model combined two transformer-based AI systems and showed high accuracy in detecting this type of hate speech. This work helps technology better recognize and moderate harmful language against men in Bengali communities.
What this means in practice
- •For content moderation teams: Detect and flag anti-male hate speech in Bengali online platforms using a specialized dataset and transformer models.
- •For natural language processing engineers: Develop and improve Bengali language models to better identify gender-based hate by incorporating a new dataset and fusion strategies.
- •For social media platform developers: Integrate models trained on the Bengali misandry dataset to enhance automated moderation against under-recognized male-targeted hate speech.$Commercial implications: Enables selling enhanced moderation tools that address gaps in gender-based hate speech detection specifically for Bengali content.
Tested on simulated data.
Authors
Md Mehedi Hasan Nipu, Chinmoy Mitra, Tarannum Ahmed Nowshin, Israt Moyeen Noumi
Abstract
Gender based hate speech research in Bengali has focused almost exclusively on misogyny, leaving anti male hate, or misandry, entirely unstudied for this language and creating a systematic blind spot in moderation systems trained to catch attacks on women but not on men. We introduce Ben-Misandry, the first Bengali misandry detection dataset: 5,000 comments across six taxonomy classes, including a novel category for mocking male emotional vulnerability with no direct analogue in existing misogyny taxonomies. The corpus is taxonomy-controlled synthetic data generated with GPT-4.1, spot-checked by four human reviewers on a 500-comment sample who confirmed the large majority of assigned labels; a full inter-annotator- agreement study is left to future work. We benchmark it with a classical pipeline (seven classifiers over four text representations) and a transformer pipeline (BanglaBERT, XLM-RoBERTa, and two fusion strategies). The best classical model, a linear SVM over character-level TF-IDF, reaches 91.8% accuracy; a late fusion ensemble averaging BanglaBERT and XLM-RoBERTa softmax outputs reaches 96.6% accuracy and macro ROC-AUC above 99%, nominally ahead of every single model and early fusion, though we have not run significance testing to confirm these margins are meaningful. We relate this to classifier combination theory, under which averaging encoders with decorrelated errors is expected to reduce variance, and corroborate the decision boundary with Shapley explanations that recover the same class- discriminative vocabulary found in our exploratory analysis, while noting this cannot rule out the model partly keying on synthetic-generation artifacts. Our results indicate misandry detection in Bengali is feasible and measurable on this synthetic corpus, offered as an initial public resource pending validation on real-world data.