PE2LGP: translating European Portuguese into Portuguese Sign Language glosses

Keywords: European Portuguese, Portuguese sign language, automatic translation, annotated corpus, gloss, natural language processing

Abstract

As the Portuguese language, the Portuguese sign language evolved naturally, acquiring grammatical characteristics different from Portuguese. Therefore, the development of a translator between the two languages consists in more than a mapping of words into signs (signed Portuguese), as it should ensure that the resulting signs satisfy the grammar of the Portuguese sign language and that the translations are semantically correct. Previous works use exclusively manual translation rules and are very limited in the amount of grammatical phenomena covered, producing merely signed Portuguese.

This paper presents the first translator from Portuguese to the Portuguese sign language, based on manual rules, but also in translation rules automatically built from a reference corpus. Given a sentence in Portuguese, the system returns a sequence of glosses with markers that identify facial expressions, spelled words, among others. The paper reports both a manual and automatic evaluation. Results show improvements in the translation quality of simple and short sentences compared to the baseline system ("signed Portuguese").

Moreover, this is the first study that deals with grammatical facial expressions, which mark interrogative and negative sentences.

Published
2021-07-10
How to Cite
Gonçalves, M., Coheur, L., Nicolau, H., & Mineiro, A. (2021). PE2LGP: translating European Portuguese into Portuguese Sign Language glosses. Linguamática, 13(1), 3-21. https://doi.org/10.21814/lm.13.1.338
Section
Research Articles