An Ontology-Based Multi-agent Argumentation System

When designing and using of Distributed Multi- Agent Systems, we often encounter the problem of conflict resolution. Argumentation theory provides a new idea for solving conflict resolution problems. Arguments play important roles during the whole argumentation process. However, the heterogeneity of distributed system makes arguments in these system vary greatly. It has become the main obstacle for Agents sharing arguments information and comparing arguments.To realize efficient automatic argumentation, we not only need Agents to share a unified argument form when interchanging arguments, but also have a demand for sharing arguments’ inference process based on an argument ontology. Argument ontologies can normalize the representation of arguments. An argument ontology constructed using Semantic Web labeling language which shared by all parts not only benefits automatic reasoning based on argument structure by Agents to enhance the efficiency of discovering attacking relations between arguments, but also contributes to sharing and reusing of argument information to enhance the efficiency of constructing arguments.Argument Interchange Format(AIF) and its realization were studied first in this article. Its applying scope and competence to solve problem were analyzed and summarized. And then, the representation of arguments with tree structure was studied. After this, the argument ontology ArgArena was proposed for the purpose of Multi-Agent automatic argumentation by extending AIF and using tree-structure arguments. Consistency checking using pellet was passed validating axioms defined in the Argument Ontology. Following this, an algorithm was proposed which enabled Agents query and construct arguments based on Argument O ntology using Prolog inference engine during argumentation process. On top of these studies, the Multi- Agent Argumentation System based on Argument O ntology ArguingO ntoArena was established based on Multi-Agent argument games model Arena, for the purpose of conflicts resolution between Agents through the study and comparison of several argumentation theories.Finally, the ArguingO nto Arena system, which was implemented by Java and the Agent developing platform Jade, could model argumentation processes built by Agents between whom belief conflicts existed. Results of experiments indicate that Agents in the ArguingO nto Arena system can query and construct arguments based on Argument Ontology, and reuse arguments already constructed which exists in argument ontologies during argumentation processes, thus the efficiency of automatic argumentation is enhanced. The conflict sources of Agents were revealed and conflicts were resolved after argumentation processes.