Evaluating the Valuable Rules from Different Experience Using Multiparty Argument Games

This paper proposes a dialectical analysis model for multiparty argument games to evaluate rules mined from different past experience, called Arena. This model transforms the multiparty argument games into two-party argument games using the ideas from the Arena Contest of Chinese KungFu, to model a dynamic process of finding a defensible argument from the different agents´ experience and to form a grid of dialectic analysis trees. For the same case, Arena enables the participating agents to propose their opinions and arguments, and provides them a platform to argue from experience in order to choose the valuable experience rule. When the training cases are more than enough, all the valuable rules about a set of databases will converge towards a stable set. We show how to choose the valuable experience rule from multiparty experience using Arena. This approach provides a new way to evaluate the experience rules mined from a database using argumentation. Arena demonstrates a fact that a combined analytical and inductive machine learning method could overcome the pitfalls associated with each separate approach.