Md. Mubasshir Naib

Multi-label Emotion Classification on Code-Mixed Text

• Developed and evaluated ML, DL, Transformer, and LLM-based models for multi-label emotion classification on Bangla-English code-mixed text across 9 emotion categories.
• Achieved 78.64% F1-score using Mistral-7B; paper submitted to IEEE SPICSCON 2026 (Under Review).
• Conducted extensive experiments with mBERT, XLM-RoBERTa, and fine-tuned LLMs for low-resource language processing.

Abusive Language Detection in Tamil

• Developed transformer-based approaches for detecting abusive Tamil text targeting women on social media as part of the DravidianLangTech 2025 shared task at NAACL.
• Implemented and fine-tuned multilingual transformer models (mBERT, XLM-R) for code-mixed and low-resource language classification.
• Paper accepted at NAACL 2025 DravidianLangTech Workshop.

Hate Speech and Misogynistic Meme Detection

• Built multimodal frameworks for detecting misogynistic memes in Chinese online content and hate speech detection in code-mixed social media data.
• Developed transformer-based models for caste and migration hate speech detection across multiple languages.
• Published three papers at the LT-EDI 2025 workshop covering hate speech, misogynistic memes, and racial hoax detection.