SignSinhala: A Hybrid Machine Learning Model for English-to-Sinhala Sign Language Translation

This paper presents SignSinhala, a hybrid machine learning framework designed to translate English text into Sinhala Sign Language. This system integrates natural language processing, neural gesture prediction, and video synthesis for sign language generation. Our novel approach incorporates grammar-aligned text preprocessing, synonym searching for out-of-vocabulary tokens, and a fingerspelling dataset leveraged to improve translation robustness. Experimental evaluation demonstrates strong semantic preservation and token filtering performance in preprocessing, laying the groundwork for effective sign generation. This research addresses a critical gap in sign language technology, fostering accessibility and inclusive communication for the deaf community.