| dc.description.abstract | Ontologies are models of representation of knowledge easily interpreted by both humans and computers. Bayesian networks are models of knowledge that work with uncertain reasoning, providing a way to treat uncertainty. The ontology summarization aims to facilitate and improve the understanding of an ontology, in order to restrict the knowledge of the domain to the most important concepts. There is a great difficulty in the construction and / or generation of Bayesian Networks, according to the literature. Most known models involve extensive manual interaction. On the other hand, there is a growing availability of ontologies that describe the knowledge of several areas. These ontologies can be applied as sources for the creation of Bayesian networks, through several models. This work presents a new model for the semiautomatic extraction of Bayesian Networks from ontologies. The differential of this work is the analysis of the relevance of the semantic aspects of the ontologies present in a developed conversion algorithm, as well as the structuring of the knowledge needed to be converted, where in this work we use the ontology summarization. The literature presents no approach that resembles this one. The text describes the theoretical basis and related works, as well as the formulated hypotheses, the developed model and the preliminary evaluation experiment. The model was implemented in a real case of Bayesian network generation for clinical cases and was integrated to a Bayesian Network editor. Three different experiments were carried out, through comparative analyzes, with specialists and questionnaires. The results indicate good possibilities in the generation of Bayesian networks, being effective when compared with manual results and advances in relation to the state of the art. The model was well received and endorsed by experts. | |
