International Workshop on Sign Language Translation and Avatar Technology (SLTAT)

10-11 Jan 2011

Located at the Federal Ministry of Labour and Social Affairs, Berlin, Germany

Thanks to all participants of the workshop and our invited speaker for their valuable contributions and to the two sign language interpreters. Special thanks to the Federal Ministry of Labour and Social Affairs for hosting this event.

You can download all abstracts and presentation slides under Programme.

If you would like to be informed on similar events, subscribe to the SLTAT mailing list.

(Foto showing all participants, the two sign language interpreters and the members of the e-government department at the Federal Ministry of Labour and Social Affairs)

Overview

Recent advances in virtual character technology and a common trend toward the 3D internet have the potential to achieve full accessibility for the deaf in the internet and to provide key tools to facilitate participation in a hearing world. Prior work on the topic of sign language translation using avatars has explored an impressive range of methods in various languages and application domains.

Future projects will profit from a joint effort of international experts to look back on the work done and identify and specify the remaining problems (or those only partially solved) in specific areas, such as symbolic translation and sign language animation. The motivation is threefold: First, discussing problems in detail will reveal potential solutions and, second, will lead to an exchange of ideas across sign language boundaries. Third, a common agreement on problems hopefully leads to a higher comparability of results and even objective benchmarks.

In this workshop, we want to focus on two core challenges: (1) symbolic translation and (2) sign language animation with avatars. We would like participants to address one of these challenges by conceiving answers to a number of questions that are listed below and additionally, submit a video for challenge 2. The format of the workshop will be single-track, organized in three stages for each challenge: (A) presentation, (B) discussion and (C) synthesis of ideas. We hope to arrive at tangible results that can be published in a volume or journal after the workshop.

During the workshop, we would like participants to give a short presentation (20 mins + 5 mins if a video is shown), covering both own work (50%) but particularly addressing the challenges (50%). Please prepare your abstract (1-2 pages, PDF or Word) accordingly.

Download Call for Abstracts as PDF.

Questions to be addressed:

CHALLENGE 1: SYMBOLIC TRANSLATION

In this area we want to discuss possible approaches for translating a spoken language sentence (in its written form) to a signed utterance representation, e.g. a sequence of glosses.

  • How much linguistics and lexicography is required?
  • How much world knowledge is required?
  • Do you use a corpus to inform the process, and how does the corpus have to look like?
  • What triggers the use of sign-specific means of expression, such as productive signs and classifier constructions, or constructed action?
  • What are good evaluation measures?
  • How much detail must the target signed utterance representation contain?

CHALLENGE 2: AVATARS

In this area we want to explore the capabilities of avatars currently used and come up with a definitive list of requirements and future research efforts to produce comprehensible and natural looking avatars.

  • What are requirements for the animation representation language?
  • Which degree of antropomorphism/photorealism is useful for signing avatars?
  • What sensory information from the user (location, gaze, ...) is needed for interactive setups?
  • What are relevant evaluation measures for avatars and how are they operationalized?

For the following set of questions we would ask the participants targeting Challenge Area 2 to submit a video that showcases these aspects using the avatar they currently use. This would allow a comprehensive comparison of state-of-the-art technology:

  • How does your avatar handle directed signs?
  • How does your avatar realize constructed action and constructed dialogue?
  • How does your avatar handle multi-party dialogue?
  • How much emotional expression can your avatar handle and in which contexts?
  • How flexible is the face/mouth and how much flexibility is required? (e.g. mouthing, visibility of teeth and tongue, puffed-up cheeks, eyebrows, wrinkles)

The workshop is partially funded by the German Federal Ministry of Labour and Social Affairs. Participation at the workshop therefore is free of charge. However, we will not be able to fund travel/accommodation. We would like to ask invitees to notify us about their participation by Dec 08, 2010. Please specify the number of people coming from your group/lab.

Contact and Submission

Please contact Alexis Heloir for questions and/or submission of abstracts:

alexis.heloir@dfki.de