Towards Developing Parallel Corpora for Portuguese and Portuguese Sign Languages
Sapienza Universitá di Roma, Ziba Khani · University of Maribor, University Press eBooks · 2023
Low-resource languages, including sign languages, are a challenge for machine translation research. Given the lack of large-scale parallel corpora, researchers must use small parallel corpora for training an automatic translation system. This article aims to address this problem by building artificial parallel corpora for Portuguese sign language in automatic translation systems. In this work, we obtained small parallel corpora of Portuguese text and Portuguese Sign Language gloss from the Metro of Porto. We used these corpora to learn grammar rules in translation between Portuguese text and Portuguese Sign language gloss. Ap- plying obtained rules to our data, we generated artificial parallel corpora for Portuguese and Portuguese sign language gloss.