Motivation

In the spirit of last year’s best theme paper at ACL (Including Signed Languages in Natural Language Processing), sign languages should become an integral and mainstream part of NLP research.

As it did for spoken languages, WMT shared tasks for signed languages could boost future research in this direction and also increase recognition.

We envision that there are benefits both for Deaf sign language users and for the research community.

For Deaf communities, the shared task could result in better access to linguistic tools, including MT, in their native languages and also improve recognition for sign languages. Contrary to popular belief, sign languages are natural languages and therefore full-fledged linguistic systems with their own grammars and lexicons. There is no single universal sign language, rather several hundred sign languages have been documented to date. In the European Union (EU), only five countries recognize their sign language(s) on a constitutional level, despite the fact that approximately half a million EU citizens are native sign language speakers. Moreover, the UN Convention on the Rights of Persons with Disabilities (UN CRPD) explicitly mentions sign language communication as a crucial right.

For the MT research community, we expect that the inclusion of sign languages in WMT shared tasks will educate researchers on sign languages and boost research on sign language translation. More concretely, the shared task will result in public benchmark data for MT systems, translations by many state-of-the-art systems and judgements of translation quality by humans. For sign languages, such resources do not exist at the moment. Besides, the NLP community already agrees that sign languages should become an integral and mainstream part of NLP research. This is evident from the fact that the best theme paper award at ACL 2021 was given to Including Signed Languages in Natural Language Processing.