Analysis of convergence and divergence in chat conversations

Abstract

A discussion between several participants is often accompanied by an exchange of information between speakers and, in the case of collaborative learning, a polyphonic interanimation is desired. They have, according to the polyphonic model, different points of view that are convergent or divergent by which one of the participants approves or disapproves of another person that is participating in the discussion. This paper presents an approach for identifying convergent or divergent utterances using neural networks and other machine learning methods in order to help people analyze a dialogue conducted in an online environment. Especially in collaborative chats used in education, this solution allows professors to identify how the participants in the discussion exchange information and how the debate evolved in time.