Automatic Schema Acquisition in a Natural Language Environment

This paper outlines an approach to schema acquisition. The approach, called ~ schema acauisition is applicable in problems solving situations and is heavily knowledge-based. Basically, learning is viewed as a fundamental part of the understanding process. Understanding a situation for which there is no existing schema involves generalizing the new event into a nascent schema. The new schema is then available to aid in future processing and can be further refined via that processing. This approach to learning is unique in several respects: it is not inductive and so is capable of one trial learning, it does not depend on failures to drive the learning process, and it is incremental and learns comparitively slowly. The learning procedure is outlined briefly with an example, a taxonomy of situations involving explanatory schema acquisition is given, and there is a brief discussion on the scope of the learning mechanism.