Toward developing a very big sign language parallel corpus

Abstract

The Community for researchers in the field of sign language is facing a serious problem which is the absence of a large parallel corpus for signs language. The ASLG-PC12 project, conducted in our laboratory, proposes a rule-based approach for building big parallel corpus between English written texts and American Sign Language Gloss. In this paper, we present a new algorithm to transform a part of English-speech sentence to ASL gloss. This project was started in the beginning of 2011 and it offers today a corpus containing more than one hundred million pairs of sentences between English and ASL gloss. It is available online for free in order to develop and design new algorithms and theories for Sign Language processing, for instance, statistical machine translation and any related fields. We present, in particular, the tasks for generating ASL sentences from the corpus Gutenberg Project that contains only English written texts. © 2012 Springer-Verlag.