A Method for Building the Quantitative and Qualitative Part of Bayesian Networks for Intelligent Tutoring Systems

The Bayesian network (BN) is an important technique to represent and infer knowledge in an Intelligent Tutoring System (ITS); however, ITSs are complex to build. Diverse authors have built BNs based on ontologies to accelerate the building process; nonetheless, they did not fully automate the process, and did not follow the ontologies standard Web Ontology Language, or simplified the final domain. This work proposes a method to build BNs based on ontologies and Wikipedia information to be employed on ITSs. The proposed method constructs the qualitative part of the BN through classes and relations of ontologies; the quantitative part is created based on frequencies, hops, and a measure of similarity between concepts of the ontology represented by Wikipedia articles. This study carried out an experiment to determine the correlation of our method against domain experts opinions; the method obtained a positive correlation of $0.647$ according to the Spearman test. The method constructs a BN to represent the knowledge in ITSs, in a similar way as experts would, supporting the automatic build of these systems.