TeamHerald@CHIPSAL 2026: Hate Speech Detection and Sentiment Analysis of Nepali Memes using Transformer-based Architectures and Ensemble Learning
Researchers have developed transformer-based models to analyze Nepali memes for hate speech and sentiment. The study focused on text extraction from memes, employing OCR and subsequent analysis with transformer architectures. Experiments showed that a decoder-only model excelled at binary hate speech detection, while a soft voting ensemble approach improved sentiment analysis performance by 15.8% in Macro F1-score. AI
IMPACT Demonstrates advanced NLP techniques for low-resource languages and multimodal content analysis.