Improved Prediction of Heuristic Configuration for Queue Priority on FaCt++ Semantic Reasoner
Descripción
With the adoption of the semantic web, interest in technologies and theory about formalizing the representation of knowledge, and automated reasoning has increased. Ontologies are concrete instances of knowledge representation and logical reasoners play an important role during the creation of ontologies, since they can find logical errors during design. One of these logical reasoners is FaCt++, which implements an optimized analytical tableaux algorithm. The specific implementation of tableaux in FaCt++ includes a set of priority queues to handle and expand the different operators that can be found during reasoning, these queues have an order for applying the different operators, which in this work will be referred as priority configurations. It was proved by the authors of FaCt++ that there is not a single priority configuration that has the best performance for all types of ontologies. Recently it was suggested that machine learning models can be successfully applied to find the best heuristic for every specific ontology. In this work is presented the process to build a machine learning model to find the best priority configuration for every ontology in detail. The model proposed in this work uses fewer features than the one shown previously and creates a simpler model with similar or better accuracy on the resulting classification.ITESO, A. C.