A Process for Semi-Automated Construction of Sign Language Dictionaries

• Marília Soares UFPB
• Beatriz Meireles UFPB
• Sandro Gonçalves UFPB
• Manuella Aschoff UFPB
• Tiago Maritan UFPB
• Valdecur Becker UFPB

Resumo

In the scientific literature, there are several solutions developed for machine translation of digital contents into sign languages. Generally, these solutions aim to reduce communication and access to information barriers for deafs in Information and Communication Technologies (ICTs). One of the main challenges of this type of system is the creation of visual representations (animations or videos) for the signs of the language, generally thousands of signs, which is a very time-consuming task. Thus, in this work, we propose a process (or methodology) for the semi-automatic construction of sign language dictionaries. This process combine automatic tasks, such as motion capture (e.g., using Microsoft Kinect) with manual tasks involving 3D animators and sign language specialists. To evaluate this proposal, a case study was developed for the Suíte VLibras platform, an open-source machine translation platform to Libras. The result was the creation of a Libras dictionary with more than 13,000 signs which is currently used in more than 1500 websites.