Model-based acquisition of inference structure

A model-directed acquisition of inference structures is introduced. It is a general knowledge acquisition system in that inference structures are the most common media among different domains. A library called OLD-WORLD is used to store a collection of inference structures, generalized from known domains and is used to guide the acquisition of new domain inference structures. Predicate calculus is used as the knowledge representation formalism. A system called MAC, which is used to acquire the inference structure of a domain, is introduced. The model used is quite different from others in that inference structures for generic tasks are taken as the key to access other domains. A comparison with other modeling approaches is given.<
>