O uso de semântica na melhoria dos métodos indutivos de aquisição automática de conhecimento.

Abstract

In this paper, we investigate how using expert-acquired semantics can improve the quality of responses provided by induction-based automatic learning algorithms. Initially we show how to introduce the use of semantics in the TDIDT family algorithms (the most known inductive algorithms). However, even with semantics, we still have a problem of a syntactic nature due to the way TDIDT algorithms represent the acquired knowledge (Decision Trees). In order to solve this problem, we present the PRISM algorithm, which solves the syntactic problem although it also suffers from the semantic problem as the TDIDT algorithms. Finally, we propose the RPRISM algorithm, which solves the syntactic problem like PRISM and semantics, using knowledge obtained from the expert.