A «Portrait» Approach to Multichannel Discourse.

This paper contributes to the research field of multichannel discourse analysis. Multimodal discourse analysis explores numerous channels involved in natural communication, such as verbal structure, prosody, gesticulation, facial expression, eye gaze, etc., and treats them as parts of an integral process. Among the key issues in multichannel studies is the question of the individual variation in multichannel behavior. We address this issue with the help of a multichannel resource "Russian Pear Chats and Stories" that is currently under construction (multidiscourse.ru). This corpus is based on a novel methodology of data collection and is produced with the help of state of the art technology including eyetracking. To address the issue of individual variation, we introduce the notion of a speaker's individual portrait. In particular, we consider the Prosodic Portrait, the Oculomotor Portrait, and the Gesticulation Portrait. The proposed methodology is crucially important for fine-grained annotation procedures as well as for accurate statistic analyses of multichannel data.