http://lattes.cnpq.br/3149506293482176; CARVALHO, Joelson Nogueira de.
Résumé:
In this work we present a machine learning system which starts from a set of objects formed by positive and negative samples of the concept to be learned. Each of these objects is structurally described through a graph based language and the knowledge is fitted into a frame that involves classes of objects and relationships. The system also presents a majoritary logic based heuristic control and allows the user's interference as a mean for improving the learning. Such type of learning system can be used in several applications. In this work, in particular, it is presented as a base tool for the system SAID, a scientific discovery support system based on machine learning.