Knowledge acquisition and corpus for argumentation-based chatbots
Lisa Andreevna Chalaguine, Anthony Hunter · UCL Discovery (University College London) · 2019
Many of the conversations we have every day involve exchanges of arguments and counteraguments. In the context of artificial intelligence and argumentation theory, such phenomena fall into the area of dialogical argumentation. Conversational agents, also known as chatbots, are versatile tools that have the potential of being used in dialogical argumentation. We can assume that a chatbot would take a particular stance in the dialogue, opposing the stance of the user. In order to succeed, the chatbot also needs to be aware of various arguments and the interplay between them. Such knowledge can be represented by a directed graph, where nodes stand for arguments and arcs symbolise conflicts between them. The chatbot must be aware of both sides of the discussion, i.e. the arguments that it can play as well as ones that the user might have, to be able to formulate convincing responses. The availability of large argument graphs for research, however, is very limited. This means that researchers do not have corpora available which hinders the development of new chatbots and limits the e↵ectiveness of existing ones. In this paper, we propose a method to acquire a large number of arguments in a graph structure using crowd sourcing. We evaluate this method in a study with participants and present a corpus which can be used for further research in computational argumentation and chatbot technologies for argumentation.