Bayesian Belief Network Model of Indirect Speech Act Theory
Xianbo Li, Huanrun Qiao, Zhicheng Ma, Diaodiao Yang, Yongtai Pan, Zhixin Ma · 2018
Naive Bayesian probability model has been employed in rational speech act and uncertain rational speech act. The requirement, all attributes must be conditionally independent of each other, is too strict for indirect speech act because the attributes do not necessarily satisfy the assumption of independent class conditions in discourse. Therefore, a Bayesian belief network model of indirect speech act theory is proposed to cancel the independent conditions. Cognitive model theory and non-parametric estimation methods are used to construct the cognitive attributes for the real world and the judgment attributes of the speaker's behavior to achieve the illocutionary act and perlocutionary act in the indirect speech act. The conditional probability table is constructed to make the model more objective and practical.