Beyond Benchmarks: Exposing the Hidden Crisis in Bangla Hate Speech Detection

2026-07-13

AI Summary

This study addresses the limited generalization of Bangla hate speech detection models in real-world scenarios, particularly their poor handling of implicit expressions, cultural context, and emoji semantics, which often leads to misclassification and over-censorship. The authors systematically evaluate six mainstream architectures—including FastText, CNN/LSTM/BiLSTM variants, and BanglaBERT—on both benchmark and multi-source merged datasets, and for the first time expose their generalization failure using external data from real social platforms. They propose an emoji-aware preprocessing strategy that preserves and enhances emoji semantics, achieving up to a 12% F1 improvement on implicit hate speech detection. Results show BanglaBERT attains 91.4% F1 on benchmark data but drops sharply to 63.4% on external samples containing sarcasm and emojis, underscoring the need for culturally sensitive, context-adaptive ethical moderation frameworks.

Abstract

The spread of hate speech (HS) across different social media platforms (SMPs) poses a major concern for online safety and ethical moderation. Automatic detection of HS remains a challenging task, especially in under-resourced languages like Bangla, due to cultural context, implicit expressions, and informal linguistic patterns. This study aimed to expose the crisis of Bangla HS detection systems by diagnosing how and why benchmark-trained models fail to identify implicit, context-dependent HS. Six architectures (FastText + CNN, FastText + LSTM, FastText + BiLSTM, BanglaBERT, BanglaBERT + CNN, and BanglaBERT + BiLSTM) were trained on benchmark datasets (about 75,000 posts) and a merged multi-source dataset (about 120,000 posts), then externally validated on an annotated dataset (about 200 posts) collected from Facebook, Twitter, and YouTube, labeled as HS and non-HS, where HS was further categorized as explicit and implicit. BanglaBERT achieved an F1-score of 91.4% on benchmark datasets but declined to 75.3% on the external set and 63.4% for implicit HS involving sarcasm and emojis. The accuracy of FastText + CNN dropped from 78.0% to 51.2% under similar conditions. Emoji-aware preprocessing improved implicit HS detection by up to 12%, whereas emoji removal caused a notable decline in performance (F1: 0.75 to 0.63). Frequent misclassifications in politically charged or satirical comments revealed over-policing risks. This study not only exposes the generalization crisis due to implicit, culturally embedded, and emoji-laden expressions but also underscores the need for developing adaptive, emoji-aware, and culturally grounded frameworks that ensure ethical moderation while preserving freedom of expression. Findings of this study provide insights for researchers, SMPs, and policymakers to design more context-sensitive HS detection systems for low-resource languages.

Problem

• hate speech detection
• implicit hate speech
• Bangla language
• emoji
• generalization crisis

Innovation

• hate speech detection
• implicit hate speech
• emoji-aware preprocessing
• cross-dataset generalization
• low-resource languages