Invented Predicates to Reduce Knowledge Acquisition

The aim of this study was to develop machine learning techniques that would facilitate knowledge acquisition from an expert by taking over the knowledge engineering task of identifying intermediate abstractions. As the expert provided knowledge the system would generalize from this knowledge and use the abstractions it learned in order to reduce the need for later knowledge acquisition. This generalization should be invoked automatically and be completely hidden from the expert manually adding knowledge. We have developed such a learning technique based on Duce's intra construction and absorption operators [Muggleton, 1990] and applied to Ripple Down Rules (RDR) incremental knowledge acquisition [Compton & Jansen, 1990]. Preliminary evaluation shows that this mixed initiative approach reduces knowledge acquisition effort by up to 50%.