Knowledge Acquisition in the Consul System

Many knowledge-based systems feature general machinery that operates on externally supplied information, These systems must solve the acquisition problem: how to represent the external knowledge, determine if it is adequate, and incorporate it into the knowledge base. As a mediator between users and interactive services, the Consul system must understand the intent and behavior of programs that perform interactive functions To Consul, understanding a function means classifying a description of it in a highly structured, prebuilt knowledge base. A special formalism has been designed in which a service builder both programs functions and describes their actions. The resulting functional descriptions are then translated and interactively classified into Consul's knowledge base by Consul s acquisition component The acquisition dialogue with the service builder will be shown to be robust with respect to the information provided by the service builder. Inference rules are automatically generated to account for discrepancies between a program's specifications and expectations derived from Consuls knowledge base.