Aquisição indutiva de conhecimento contemplando os aspectos sintático, semântico, de generalização e custo.

Abstract

The automatic acquisition of knowledge arose with the objective of solving the problems presented by the cognitive and semi-automatic methods, which range from the relationship of the knowledge engineer with the expert to the unavailability of time of the expert. However, the knowledge base assessment generated by the automatic methods is a process that is still done exclusively by the expert. The inductive methods of knowledge acquisition from examples, one of the most used forms of automatic acquisition, can present two sine problems, one of structural nature (syntactic problem) and another of semantic nature (semantic problem). These methods are also called empirical emphasizing the fact that they require no preliminary knowledge of the domain. Choosing an appropriate form of knowledge representation It eliminates the syntactic problem and the use of one of the forms of preliminary knowledge, called the relevance matrix, makes the semantic problem rare. The other forms of preliminary knowledge available (cost and generalization) assist inductive methods in finding a better quality knowledge base. However, no inductive method jointly utilizes the main forms of preliminary knowledge available. In order to fill this gap, in this paper we propose the ISREG algorithm (Semantic Inductor of Economic and Generalized Modular Rules) that further solves the syntactic problem and minimizes the occurrence of the semantic problem. We also propose an automatic process for evaluating the semantic quality of knowledge bases generated by inductive methods.