This upload is the accepted version of the paper. For the closed access published version follow the URL in the citation.

Cite as

M. Kopf, R. Omardeen and D. Van Landuyt, "Representation Matters: The Case for Diversifying Sign Language Avatars", 2023 IEEE International Conference on Acoustics, Speech, and Signal Processing Workshops: Sign Language Translation and Avatar Technology, Rhodes Island, Greece, 2023, doi: 10.1109/ICASSPW59220.2023.10193409. URL: https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10193409

Abstract

As interest in sign language machine translation grows, there is more focus on developing and refining avatar technology to present translation outputs. While signing avatar technology research has focused on legibility and appearance, there has been little attention paid to representing diversity in signing avatars, and the default is often a white female animation. We present data from focus groups conducted in two ongoing sign language machine translation projects in Europe that give insight into deaf end-users' desires for diversity in avatar representations. Our results reveal a strong desire for full customisability, including options for representing diversity in gender expression and ethnicity, as well as accommodating sociolinguistic variation and personal identity through modified avatar signing styles. This work provides initial insights, but considerable future research is necessary, particularly with minorities and sub-groups within deaf communities.

©2023 IEEE