Gradient Masters at BLP-2025 Task 1: Advancing Low-Resource NLP for Bengali using Ensemble-Based Adversarial Training for Hate Speech Detection
This work addresses hate speech detection for Bengali, a low-resource language, but is incremental as it builds on existing language models and ensemble techniques.
The paper tackled hate speech detection in low-resource Bengali by developing an ensemble-based adversarial training method, achieving a micro F1 score of 73.23% for hate-type classification and 73.28% for target group classification in a shared task.
This paper introduces the approach of "Gradient Masters" for BLP-2025 Task 1: "Bangla Multitask Hate Speech Identification Shared Task". We present an ensemble-based fine-tuning strategy for addressing subtasks 1A (hate-type classification) and 1B (target group classification) in YouTube comments. We propose a hybrid approach on a Bangla Language Model, which outperformed the baseline models and secured the 6th position in subtask 1A with a micro F1 score of 73.23% and the third position in subtask 1B with 73.28%. We conducted extensive experiments that evaluated the robustness of the model throughout the development and evaluation phases, including comparisons with other Language Model variants, to measure generalization in low-resource Bangla hate speech scenarios and data set coverage. In addition, we provide a detailed analysis of our findings, exploring misclassification patterns in the detection of hate speech.