The "Corpus of Interactional Data" (CID) - Multimodal annotation of conversational speech"

The understanding of language mechanisms needs to take into account very precisely the interaction between all the different domains or modalities, which implies the constitution and the development of resources. We describe here the CID (Corpus of Interactional Data), an audio-video corpus in French recorded and processed at the Laboratoire Parole et Langage (LPL). The corpus has been annotated in a multimodal perspective including phonetics, prosody, morphology, syntax, discourse and gesture studies. The first results of our studies on the CID lead to confirm the relevance of an analysis which takes into account as many linguistic fields as possible to draw up a more precise knowledge of discourse phenomena. MOTS-CLÉS : schéma d'encodage multimodal, outils et plate-forme d'annotation, phonétique, prosodie, morphologie, syntaxe, discours, geste.