Sign Languages and Machine Translation: Challenges and Opportunities

Machine translation of spoken language has benefitted from significant technological advances in recent years, owing to a large amount of scientific research and development within the domain. Unfortunately, language technology (LT) has not advanced at nearly the same pace within the field of sign language machine translation (SLMT), excluding deaf and hard-of-hearing individuals, while further suppressing a set of minority languages that are already under-resourced. This chapter sets out some of the challenges that must be addressed to foster progress within the field of SLMT. We consider this from both a linguistic and technical perspective. We begin by providing an account of the challenging linguistic phenomena associated with sign languages, that have not been sufficiently accounted for to date, due to under-resourcing. For this chapter, we take Flemish Sign Language (VGT) as our sign language of focus, but the under-described linguistic phenomena set out here hold equally for most sign languages across the world. We then consider some of the technical challenges that arise in relation to processing and modelling sign language. We provide an overview of the varying approaches taken within the field of SLMT. We outline some of the technical challenges, which must be overcome if we are to advance modern LT within the field.