The CID Video Corpus: A Multimodal Resource for Gesture Studies - Archive ouverte HAL Accéder directement au contenu
Communication Dans Un Congrès Année : 2007

The CID Video Corpus: A Multimodal Resource for Gesture Studies

Roxane Bertrand
Philippe Blache

Résumé

The CID video corpus is a project currently under development at the Laboratoire Parole et Langage (CNRS) at Aix en Provence. The ultimate aim of the project is to build a platform for the semi-automatic annotation of video corpora and share the video files as well as their corresponding annotations with the rest of the scientific community. At present, the database is composed of a video corpus of 5 hours of spoken French (dialogic conversations - 10 speakers). The audio files have been transcribed in an enriched orthographic transcription and annotated automatically at the prosodic level in Praat using the INTSINT/MOMEL algorithms developed at Aix. It has then been automatically annotated at the morphosyntactic and syntactic levels with the syntactic analyzer proposed by VanRullen (2004). We also consider the pragmatic level for specific phenomena like discourse markers, backchannel signals as well as larger discourse units like turn taking. We are now annotating the gesture level using Anvil. For this particular annotation, we use an enriched version of the gesture description proposed by McNeill (1992) and the MUMIN coding scheme (Allwood et al., 2005), adding to the transcription the annotation of trunk positions. We also revised the already quite complete Mumin coding scheme so that levels of annotation are not redundant: whereas in Mumin, the gestures are transcribed according to the speaker's role in the interaction, in our scheme, this role is transcribed independently from the gesture annotation. It will still be eventually possible to retrieve the information on interaction with an XQuery program developed conjointly at Toulon and Paris at the LIMSI and LSIS laboratories which will allow complex queries through multiple XML files. In the end, the project will not be limited to the transcription of French, tools being customized for other languages as well such as English.

Domaines

Linguistique
Fichier non déposé

Dates et versions

hal-00666074 , version 1 (03-02-2012)

Identifiants

  • HAL Id : hal-00666074 , version 1

Citer

Gaëlle Ferré, Roxane Bertrand, Philippe Blache. The CID Video Corpus: A Multimodal Resource for Gesture Studies. Third ISGS conference 'Integrating Gestures', 2007, Evanston, Illinois, United States. ⟨hal-00666074⟩
682 Consultations
0 Téléchargements

Partager

Gmail Facebook X LinkedIn More