%0 Journal Article
%T HuComTech Multimodal Corpus Annotation
%A P¨˘pay
%A Kinga
%A Szeghalmy
%A Szilvia
%A Szekr¨¦nyes
%A Istv¨˘n
%J Argumentum
%D 2011
%I Debrecen University Press
%X The Hungarian audio-visual corpus recording and annotation project is being carried out by the HuComTech (Hungarian Human-Computer Interaction Technologies) research group at the University of Debrecen and is a part of a comprehensive multimodal human-machine interaction modelling project. The research contributes to the exact knowledge of the overlaps between the verbal and nonverbal aspects of communicative events and prosodic features through the examination of spontaneous speech, with special regard to syntactical embeddings, insertions, iterations, hesitations and restarts, various kinds of emotions and discourse markers. The efficiency of speech recognition systems can also be increased by proper acoustic preprocessing and investigation of the suprasegmental characteristics of spontaneous speech. Concerning Hungarian, the lack of a multimodal, prosodically labelled, representative spontaneous speech corpus makes the development more difficult. The spontaneous multimodal corpus is being recorded via guided ¨C formal and informal ¨C conversations. During the conversations, several points are to be discussed in order to provoke longer monologues accompanied by gestures, facial expressions, and also including the above mentioned phenomena of spontaneous speech to be examined. The present paper focuses on the aspects of multimodal annotation, especially the details of the annotation of prosodic and suprasegmental features. The visual, nonverbal channel of these phenomena are also to be annotated thus we can examine and implement multimodal features and their overlaps as well.
%K Hungarian audio-visual corpus recording
%K audio annotation
%K video annotation
%U http://argumentum.unideb.hu/2011-anyagok/works/PapayK_SzeghalmySz_SzekrenyesI.pdf