Abstract

This paper presents a comprehensive study of deep learning techniques applied to natural language understanding tasks in low-resource languages. We propose a novel transfer learning framework that leverages multilingual pre-trained models and demonstrates significant improvements over existing baselines across multiple Indian languages. Our approach achieves state-of-the-art results on sentiment analysis, named entity recognition, and text classification tasks with limited training data.

Keywords

Citation

QR Code