In this thesis, we propose a multitask learning based method to improveNeural Sign Language Translation (NSLT) consisting of two parts, a tokenizationlayer and Neural Machine Translation (NMT) The tokenization part focuses on how Sign Language videos should be represented to be fed into the otherpart . We aim to develop a generic sign-level tokenization layer sothat it is applicable to other domains without further effort . We succeed in enabling knowledge transferbetween SLs and improve translation quality by 5 points in BLEU-4 and 8 points in ROUGE scores . Apart from these, we adopt3D-CNNs to improve efficiency in terms of time and space . Lastly, we discussthe advantages of Sign Language tokenization over gloss-leveltokenization over the advantages of sign-to-texting over the tokenization . To sumup, our proposed method eliminates the need for gloss level annotation to obtain higher scores by providing additional supervision by utilizing weaksupervision sources. Apart fromThese, we adopted3D CNNs to improved efficiency

Author(s) : Alptekin Orbay

Links : PDF - Abstract

Code :

https://github.com/tensorflow/tensor2tensor




Keywords : sign - tokenization - translation - language - points -

Leave a Reply

Your email address will not be published. Required fields are marked *