A debating ontology for argumentative dialogue systems

This thesis designs and implements an ontology which encodes the argumentative structure of natural language debates. This Debating Ontology will constitute the database for a prospective Argumentative Dialogue System, whose artificial agent conducts spoken language debates against human opponents on a specific subject, drawing from the argument structure encoded in the ontology. We use methods of argumentation mining to break down a written sample debate into its argument components and relations employing annotation techniques. To this end, we devise an argument annotation scheme as well as detailed annotation guidelines. In addition, we extract a set of keywords from the debate's argument components which will be utilized for the Natural Language Understanding of arguments via similarity matching. We also provide guidelines for said keyword extraction. We design and implement the Debating Ontology, test it and evaluate the quality of the argument annotation and extracted keywords. Both evaluations yield promising results