Text-to-ISL Gloss Translation: Bridging Language Accessibility Through NLP

The lack of representation of Indian Sign Language (ISL) in technological innovation hinders communication for India's Deaf community. This research addresses this problem by introducing a system for transcribing English sentences into ISL gloss, a formalized textual representation of ISL grammar. The proposed approach utilizes a Sequence-to-Sequence (seq2seq) model with a Bidirectional LSTM (BiLSTM) encoder and an LSTM decoder to capture the mapping of English syntax to ISL gloss structure. Due to the absence of public English-ISL gloss datasets, a specialized corpus of 10368 sentence pairs was created in collaboration with ISL experts. The system's performance was evaluated using BLEU and ROUGE-L F1 scores on a test set of 2074 sentences, yielding encouraging results under various training settings. This work provides a foundation for future studies in ISL video generation and its integration into accessible education and communication devices, ultimately aiming for enhanced digital accessibility and inclusive technology.