Translation of Spanish Text to Mexican Sign Language Glossed Text Using Rules and Deep Learning

People with hearing disabilities face various communication problems in their environment, both social, workplace, and educational. The World Health Organization reports that more than 360 million people worldwide suffer from hearing problems. The number of certified Mexican Sign Language (MSL) interpreters is low, with approximately 40 in Mexico, which has the consequence of not having access to higher education in this community. This article presents an architecture to translate from Spanish to MSL glossed text. This translation architecture obtained results through various models evaluated with the BLEU, WER, and accuracy metrics, confirming the viability with traditional and computational learning models for sign language translation. Translation with traditional techniques was found to perform better than deep learning. In addition, it was possible to build a corpus of ten sentences written in gloss, validated by a member of the deaf community of MSL.