Quantitative Argumentation Model Combining Content Analysis Of Argument Texts And Its Application

Debate is one of the ways to solve complex decision problems.Argumentation model is a formal description of the argumentation,and its purpose is to find the set of acceptable arguments.Most of the existing argumentation models are based on Dung's abstract argumentation framework.Its features are to ignore the content of argument texts,abstract the arguments into a node,consider only the attack or support relationship between the arguments,and obtain the final result by solving the semantic extension.Some argumentation models consider the quantitative representation of the argument and the attack intensity between arguments,ultimately solving the credibility of arguments.But still ignored the content of the argument text.In practical debates,the relationship between arguments is often hidden in the content of the argument text.When there are multiple experts who propose similar content views on the same issue,if the argument evaluation algorithms are used for calculation without processing,there will be cases where the credibility of the disputed node is repeatedly calculated.In order to solve the problems above,this paper proposes a quantitative argumentation model that combines content analysis of argument text.It presents the algorithm for the reduction of argument nodes and the argument evaluation algorithm,and applies them to the online argumentation platform.The specific work is as follows:(1)Combining the current research status and actual argument application requirements,a quantitative argumentation model that combines the content analysis of argument texts is proposed.The model is based on the IBIS model,which is based on the abstract debate.Firstly,the abstract debate graph is quantified and expressed using the credibility method.And the uncertainty of the argument and the attack intensity between the arguments are unified by the credibility factor;Then the paper analyzes the text content of the argument node,identifies the similarity of the content,and uses the argument node reduction algorithm to aggregate the argument nodes with similar content.Finally,a credibility-based dispute evaluation algorithm was used to calculate the credibility of each dispute,and the consensus value of the solution is ranked according to the credibility value,and the final acceptable solution set is obtained.(2)A reduction algorithm based on the content analysis of the argument is proposed.The algorithm is divided into two parts,the argument node text clustering algorithm and the argument node aggregation algorithm.First of all,from the root of the debate tree,text clustering is performed for each node's sibling nodes,so that nodes with similar text content are clustered into clusters.Then according to the clustering results,the nodes with similar text content are aggregated into one node to prune the debate tree and correct the credibility of the relevant disputed nodes.(3)A credibility based argument evaluation algorithm is proposed to calculate the credibility of argument nodes.The algorithm is divided into two parts,the credibility synthesis algorithm and the credibility transfer algorithm.The reliability synthesis algorithm is used to synthesize the comprehensive response strength of the sub argument nodes,and the reliability transfer algorithm is used to calculate the influence of the credibility updating of the sub argument nodes on the ancestral argument nodes.Finally,according to the argumentation model provided in this paper,we designed and implemented a debate subsystem based on B/S framework,and integrated it into the whole research system,and proved the effectiveness of the model through a concrete example.