Prosody and gestures to modelling multimodal interaction: Constructing an Italian pilot corpus

Abstract

Modeling dialogue implies detecting natural interaction. A pragmatic approach allows to consider the linguistic act composed of several and different features interacting with each other. Data collected for this project comprises three different genres of communication: monological, dialogical and conversational. The project aims to identify and analyze the pragmatic value of multimodal communication spotting the linguistic actions which carry out illocution values. We draw a pragmatic approach to study multimodal interaction combining the L-AcT annotation (Cresti 2000) with the gesture's architecture designed by Kendon (Kendon 2004). The annotation system is designed to divide the speech units (utterance, intonation units and illocution types) (Hart, Collier, and Cohen 2006) (Cresti 2005) (Moneglia and Raso 2014) from gestural units (Gesture Unit, Gesture Phrase, Gesture Phase). Keeping the Gesture Unit as a superior macrounit at the other gestural units only for the quantitative purpose, we realize a matching between gesture and speech units. These units work together to form the communicative intention of the speaker that can be recognizable by the Illocution Type. This annotation system leads to understanding how speakers realize multimodal linguistic actions and how different modalities work.