Real Time English to British Sign Language Translation for Accessible Banking

Abhishek Bharadwaj Varanasi, Manjira Sinha, Tirthankar Dasgupta, Charudatta Jadhav

Abstract: This paper presents a mobile-based framework for real-time English text-to-British Sign Language (BSL) translation, designed for accessible banking. The system uses a linguistically informed transformer model to translate English sentences into BSL gloss sequences, which are then converted into HamNoSys for avatar animation. The framework is optimized for handheld devices, enabling applications like bank notifications for the Deaf community. To support experimentation, an English-BSL parallel dataset was created, covering multiple domains. The proposed model achieved a 87.31% ROUGE-L score for text-to-gloss translation, demonstrating its effectiveness across domains. The system outputs were subjected to a user-based evaluation with BSL experts, providing valuable qualitative insights.