Study On An Argumentation Model Based On Evidence Theory

Argumentation is a conversational method for resolving the conflict. In artificial intelligence, argumentation model is used for modelling daily argumentation to representing the process of argumentation so that this model is used in non-monotonic reasoning and knowledge base maintenance. Due to the uncertainty of real arugmentation and the abstrusity of semantic extension for argument evaluation in abstract argumentation framework, the uncertain reasoning method need to be introduced into argumentation model. The argument evaluation based on uncertain reasoning is a value computing method. DS evidence theory is a widely used uncertain reasoning which can representing the ignorence well. The information can be combined by Dempster rule in DS evidence theory. The argumentation model can be more informal for real world by being improved by evidence theory. And, the argumentation model would be more practical by replacing the semantic extension by value computation.The abstract argumentation framework and the extension argumentation framework are developed in our work. The argument is made up of premise and conclusion and the relation of arguments is represented by respond of premise and conclusion. The definitions in argumentation framework combined with DS evidence theory are redefined by uncertainty part so that the argumentation model is more applicable in real world. The opinion, a triple, is proposed to represents the belief,disbelief and ignorence of the statement in argument. And, the rule strength is proposed to represents the strength of the relation between arguments. The combination and updating methods of opinion is based on reasoning based on evidence theory. The opinion upadating algorithm is used to upating all opinions of statements in an argumentation process. All statements would be analysed by opinion and all arguments would be devided into three sets: acceptable set, plausible set and unknown set. The arguments evaluation is baesd on these three sets.Finally, an example is given to illustrate the validity of the method. The conclusion illustrates that the argumentation is modeled by our argumentation model correctly and the uncertainty in argumentation is represented quantificationally so that the acceptability of arguments can be computed and the argumentation may evaluated more reasonable.