Year: 2025
Type: article
Abstract: 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.
Authors: Shanmukhaa M S -, Geetha Ramani R -
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W4410431507