Co-designing a Sign Language Avatar for Railway Announcements
Keywords
Loading...
Authors
Issue Date
2022-04-21
Language
en
Document type
Journal Title
Journal ISSN
Volume Title
Publisher
Title
ISSN
Volume
Issue
Startpage
Endpage
DOI
Abstract
Public transport organisations, such as Dutch railway operator Nederlandse Spoorwegen
(NS), generally communicate important messages through spoken announcements
in the train or station. As a result, Deaf people miss crucial information.
Because of varying levels of reading pro ciency, textual information is not always
a solution. Moreover, video footage is not scalable and will not yield satisfactory
results when individual video parts need an update. Virtual sign language avatars
can provide consistent, anonymous and scalable translations. For the development
of a comprehensible and
uent sign language avatar, collaboration between experts
in language technology, avatar technology and sign language is fundamental. Unfortunately,
inclusive approaches are not evident in existing technologies.
To address these problems, we present a case study on co-designing a sign language
avatar for the automatic translation of railway announcements into Sign Language
of the Netherlands (NGT). For the initial design, video sign language translations of
NS announcements created by the Dutch Sign Center were annotated and used for
the translation basis. A scripted keyframe animation technique for the JASigning
avatar engine makes it possible to e ciently create many variants of a given template,
without expensive equipment. Three iterative co-design sessions took place
with an interdisciplinary research team including deaf sign language experts. Simultaneously,
hearing passengers from NS were consulted for additional remarks on
the sign language avatar. Subsequently, multiple phrase variants of the system were
evaluated within a diverse focus group audience to account for demographic di erences.
Combining the various disciplines led to mayor adjustments of the avatar
(manual movements, facial expressions, mouthing, grammar, transitions between
signs, camera angle and speed). With this case study, we aimed to e ectively and
inclusively develop sign language avatar technology. More research and focus groups
are necessary to ensure high quality translations (improved mouthings, facial expressions
and prosody), resulting in enhanced comprehensibility and natural appearance
of the avatar.
Keywords: machine translation, signing avatar, co-design, focus group, railway announcements
Description
Citation
Faculty
Faculteit der Sociale Wetenschappen