Embedded Files
SUMMARY:
SUMMARY:
The Second International Workshop on Automatic Translation for Sign and Spoken Languages (AT4SSL 2023) was hosted by the 24th Annual Conference of The European Association for Machine Translation (EAMT 2023). The main theme of the 2023 edition of the AT4SSL workshop was Sign language parallel data - challenges, solutions and resolutions: Data is one of the key factors for the success of today's AI, including language and translation models for sign and spoken languages. However, when it comes to processing sign language and training machine-learning systems we face the problems of small volumes of (parallel) data, large veracity in terms of origin of annotations (deaf or hearing interpreters), non-standardized annotations (e.g. glosses differ across corpora), video quality or recording setting, and others. In this edition of the workshop we focus on the discussion of data quantity, data quality, (re)sources, ethical and ownership concerns.
AT4SSL 2023 received a total of 9 new submissions (4 long and 5 short papers), 6 of which were accepted (3 long and 3 short papers). The accepted papers cover a diverse range of topics related to automatic translation between signed and spoken languages, and focus on data resources, linguistics and machine translation (MT) systems.
The papers by McGill, E. and Saggion, H., and by De Sisto, M. et al. present new sign language corpora for BSL, VGT and NGT
The paper by Declerck, T. et al. presents a unified RDF-based representation for various type of data aiming to facilitate a common signed and spoken language repository
Moryoseff, A. et al. present a new baseline MT based on a transformation from text to glosses to poses and to video in the context of translation into SL for DGS.
The work presented by the paper of Hollain, N. et al. investigates the use of approximative linguistic features for sign language processing, seeking improvements over landmark-based features
The paper by Mohammed, Z. and Murtagh, I. presents work on integrating gerunds of Irish SL in a computational lexicon framework.
SignON and EASIER:
SignON and EASIER:
AT4SSL 2023 was organised jointly by members of the SignON and EASIER projects. Both SignON and EASIER are Horizon 2020 projects, funded under the Horizon 2020 program ICT-57-2020 - "An empowering, inclusive, Next Generation Internet" with Grant Agreement number 101017255 and 101016982 respectively.
Page updated
Report abuse