The design of traditional corpora for linguistic analysis aims to provide living representations of sign languages across deaf communities and linguistic researchers. Most of the time, the sign language data is video-recorded and then encoded in a standardized and homogenous structure for open-ended analysis (statistical or phonological studies). With such structures, sign language corpora are described and annotated into linguistic components, including phonology, morphology, and syntactic components. Conversely, motion capture (MoCap) corpora provide researchers the data necessary to carry on finer-grained studies on movement, thus allowing precise, and quantitative analysis of sign language gestures as well as sign language (SL) generation. One the one hand, motion data serves to validate and enforce existing theories on the phonologies of sign languages. By aligning temporally motion trajectories and labelled linguistic information, it thus becomes possible to study the influence of the movement articulation on the linguistic aspects of the SL, including hand configuration, hand movement, co-articulation or synchronization within intra and inter phonological channels. On the other hand, generation pertains to sign production using animated virtual characters, usually called signing avatars. Although MoCap technology presents exciting future directions for sign language studies, tightly interlinking language components and signals, it still requires high technical skills for recording, post-processing data, and there are many unresolved challenges, with the need to simultaneously record body, hand motion, facial expressions, and gaze direction. Therefore, there are still few MoCap corpora that have been developed in the field of sign language studies. Some of them are dedicated to the analysis of articulation and prosody aspects of sign languages, whereas recent interest in avatar technology has led to develop corpora associated to data-driven synthesis. This paper describes four corpora that have been designed and built in our research team. These corpora have been recorded using MoCap and video equipment, and annotated according to multi-tiers linguistic templates. Each corpus has been designed for a specific linguistic purpose and is dedicated to data-driven synthesis, by replacing signs or groups of signs, by composing phonetic or phonological components, or by altering prosody in the produced sign language utterances.
@InProceedings{GIBET18.18020, author = {Sylvie Gibet}, title = {Building French Sign Language Motion Capture Corpora for Signing Avatars}, booktitle = {Proceedings of the Eleventh International Conference on Language Resources and Evaluation (LREC 2018)}, year = {2018}, month = {may}, date = {7-12}, location = {Miyazaki, Japan}, editor = {Mayumi Bono and Eleni Efthimiou and
Stavroula-Evita Fotinea and Thomas Hanke and
Julie Hochgesang and Jette Kristoffersen and
Johanna Mesch and Yutaka Osugi}, publisher = {European Language Resources Association (ELRA)}, address = {Paris, France}, isbn = {979-10-95546-01-6}, language = {english} }
Summary of the paper