Construction of conditional probability tables of Bayesian networks using ontologies and Wikipedia

Building Bayesian Networks automatically serves to reduce time and effort by defining variables and the quantitative relation between variables. However, the quantitative part is the most complicated to solve because of it is statistical information. This research proposes a method to construct the quantitative part of a Bayesian Network based on text mining and ontologies for Intelligent Tutoring Systems. The network structure is built based on the variables and relations of an ontology. Conditional Probability Tables (CPT) are created from Wikipedia information. The constructed CPT reach a correlation of 0.895 against the experts' opinion. This correlation is good due to the subjectivity in the evaluations. We conclude that using the text mining in Wikipedia and ontologies, it is possible to construct CPT that adequately represents knowledge in an educative environment.