Uma arquitetura genérica para tradução automática text-to-gloss baseada em regras

Abstract

Deaf people have di culties in communication and information acquisition, because in both cases what is prevalent is the use of oral languages, but the deaf have sign language as natural language, having oral language as a secondary language. Several areas are seeking solutions to combat this barrier with the deaf community. In the area of technology, e orts are focused on the accessibility of Information and Communication Technologies (ICTs). One of the main strategies is the construction of automatic translators that translate oral language information into sign language. As an aid to this area of research, this paper aims to present a exible architecture proposal for the construction of automatic rules-based translators for translations between oral and sign language, based on the automatic translator used in the Su te VLibras, a consolidated system that allows the accessibility of audio, text and video in Portuguese for users of LIBRAS.As proof of concept of the proposed architecture, an automatic translator for the translation of English text for ASL gloss was created. Automatic translation tests were applied on this translator and the results were evaluated with the BLEU and WER metrics.