A Learning method based on partial structures of explanations

The learning process of macro-operators is efficient since it is guided by given operators. By extending this method, a learning method of macro-operators is proposed based on partial structures of explanations. Although until now macro-operators have been used only for deductive learning that improves execution speed, the proposed method also acquires new knowledge by using hypothesis generators, i.e., it uses the explanations involving hypothesis generators to carry out induction. We demonstrate deductive and inductive aspects of our method by an example of logic circuit design and English-Japanese translation.