Disagreement Dialogue

This paper has focused on discovering multimodal compliance/dissent patterns based on data from the HuComTech Corpus. It advocates for a multimodal approach to human interaction by highlighting the interdependence of text, language and gestures in communication, and shows how important it is to implement human behaviours in more user-friendly human-machine interaction systems. It describes the main features of body annotation with an emphasis on the classes primarily responsible for expressing consent/disagreement. After a brief introduction to the basics of the Theme search environment, relevant behavioral models of different complexities discovered by Theme were presented. “default_disagree” = an element of the contract class, i.e. a default opinion event. The corpus is the result of joint efforts of researchers in computer science, pragmatism, engineering and information sciences and psychology. Launched in 2009, the project aimed to conduct a detailed study of human-human communication in order to make an important contribution to the establishment of different human-computer interaction systems. From the beginning, it became clear that such a system should be multimodal, that is, it should go beyond verbal communication and include gestures that would increase the ease of use of these systems. It was also clear that the system should be able to model two-way communication. Indeed, it should be able to participate in a recursive series of interaction events by going beyond simply responding to a request or executing a request – it should “listen” to the reactions of the human user, evaluate them and act accordingly.

Such a system requires two simultaneously active communication channels, that of analysis and synthesis, which allows actors to continuously change their roles as spokespersons and auditors. The model we proposed as the basis of our body building followed precisely this requirement (cf. Hunyadi, 2011). The natural nature of reciprocal communication necessarily presupposes that the actors are freely involved in the given theme and that the flow of interactions allows a full and complete expression of gestures and emotions. As a result, we designed two types of dialogues: one essentially formal – in the form of a job interview with a series of predefined twists and turns and a second that constitutes an informal conversation (for better data management, the latter followed a certain guide, but allowed for individual variations). . . .