Not taking into account budget restrictions, the setup of a sign language studio always is a balancing act between high quality recordings on the one hand not to make the transcription process even more complicated than it is anyway and possibly to enable automatic processing of the recordings, and on the other hand an environment where the informants still feel comfortable enough so that the recording situation does not have too much impact on the signing. In the case of the DGS Corpus project, an additional constraint is that the studio is to be relocated twelve times over the course of two years as it was decided to make the recordings in the regions instead of inviting participants to one central place to avoid dialectal mixing. One of the implications of this approach is that the studio is operated by non-specialist deaf fieldworkers with limited time available for training. Basically all tasks in the project involve two informants interacting in different ways with each other. A moderator (the fieldworker from the region) introduces the tasks and observes the conversation, but only interferes with the conversation if absolutely necessary. The camera setup we finally ended up with consists of seven cameras altogether, three on each informant and one for the whole scene including the moderator. Two HD cameras on the informant provide frontal and birds-eye views while a stereo camera mounted on top of the frontal-view camera provides footage that helps automatic processing. The seventh camera is an HD camera as well. In contrary to setups in earlier projects, we invite the two informants to sit down directly facing each other, with the frontal-view camera positioned above (and behind) the head of the other informant. Pre-tests revealed that with a distance of approximately three meters, the distorsion introduced by the elevated position of the camera does not negatively affect the transcription from video. Instead, this setting provides a front view of the informant similar to the addressee's, allowing to identify body shifts as well as direction of eye gaze more easily. At the same time, this constellation avoids informants targeting their signing back and forth between the addressee and the camera. Elicitation material and instructions are presented to the informants on screens located on the floor between them. A custom software, “Session Director” allows the moderator to present slides to the informants by the click of a button, and to keep track of the time elapsed for each individual task as well as the whole session. Using pre-recorded instructions and elicitation materials not only reduces the burdens on the moderator, but also makes sure that all informants get exactly the same input. Session Director keeps a log of all actions started by the moderator, allowing us to exactly reconstruct what task has been worked on when. This log is easily converted into tagging in our transcription environment, iLex. This not only allows automatic segmentation of tasks and pauses, but also introduces links from the transcript to the task and vice versa. Task descriptions for Session Director are kept as XML files, making it easy to use this freely available tool for other projects as well.
@inproceedings{hanke:10047:sign-lang:lrec,
author = {Hanke, Thomas and K{\"o}nig, Lutz and Wagner, Sven and Matthes, Silke},
title = {{DGS} {Corpus} {\&} {Dicta-Sign}: The {Hamburg} Studio Setup},
pages = {106--109},
editor = {Dreuw, Philippe and Efthimiou, Eleni and Hanke, Thomas and Johnston, Trevor and Mart{\'i}nez Ruiz, Gregorio and Schembri, Adam},
booktitle = {Proceedings of the {LREC2010} 4th Workshop on the Representation and Processing of Sign Languages: Corpora and Sign Language Technologies},
maintitle = {7th International Conference on Language Resources and Evaluation ({LREC} 2010)},
publisher = {{European Language Resources Association (ELRA)}},
address = {Valletta, Malta},
day = {22--23},
month = may,
year = {2010},
language = {english},
url = {https://www.sign-lang.uni-hamburg.de/lrec/pub/10047.pdf}
}
| König, Lutz