Content Determination for Natural Language Descriptions of Predictive Bayesian Networks

The dramatic success of Artificial Intelligence applications has been accompanied by more complexity, which makes its comprehension for final users more difficult and damages trustworthiness as a result. Within this context, the emergence of Explainable AI aims to make intelligent systems decisions and internal processes more comprehensible to human users. In this paper, we propose a framework for the explanation in natural language of predictive inference in Bayesian Networks (BN) to non-specialized users. The model uses a fuzzy syllogistic model for building a knowledge base made up of binary quantified statements that make explicit in a linguistic way all the relevant information which is implicit in a BN approximate reasoning model. Through a number of examples, it is shown how the generated explanations allow the user to trace the inference steps in the approximate reasoning process in predictive Bayesian Networks.

keywords: Content Determination in natural language generation, Linguistic descriptions, Fuzzy syllogism.