Spanish generation from Spanish Sign Language using a phrase-based translation system

Resumen

This paper describes the development of a Spoken Spanish generator from Spanish Sign Language (LSE – Lengua de Signos Española) in a specific domain: the renewal of Identity Document and Driver’s license. The system is composed of three modules. The first one is an interface where a deaf person can specify a sign sequence in sign-writing. The second one is a language translator for converting the sign sequence into a word sequence. Finally, the last module is a text to speech converter. Also, the paper describes the generation of a parallel corpus for the system development composed of more than 4,000 Spanish sentences and their LSE translations in the application domain. The paper is focused on the translation module that uses a statistical strategy with a phrase-based translation model, and this paper analyses the effect of the alignment configuration used during the process of word based translation model generation. Finally, the best configuration gives a 3.90% mWER and a 0.9645 BLEU.