From Speech and Gestures to Dialogue Acts

In the present paper, selected problems related to the annotation of the DiaGest corpus are discussed. A system of dialogue acts is proposed along with a conceptual framework that allows for independent labelling of the contributions provided by various modalities and channels. Both auditory and visual modalities are considered. Four channels are defined as major ways of providing quasi-independent modal contributions. A four-dimensional categorisation of dialogue acts is proposed. It includes a separate dimension for attitude-related tags. Dialogue acts are conceptualised as multidimensional entities built on the basis of modal contributions provided by respective channels.