Learning Causal Semantic Representation from Information Extraction

For reasoning with uncertain knowledge causal semantic analysis is proposed to construct logical rules,which are extracted from decision tree induction and Bayes inference based on generalized information theory. These rules can represent multi-level semantic knowledge of the relationship between the data and information implicated. Empirical studies on a set of natural domains show that the semantic completeness of generalized information theory has clear advantage in representing semantic knowledge from different levels.