Interview-based knowledge acquisition using dynamic analysis

The use of domain- and task-independent primitives to facilitate writing task-specific interview strategies is discussed. Using both static and dynamic analysis makes it possible to begin knowledge acquisition with an incomplete domain model and later refine and build the knowledge base. Knowledge can be acquired for cases not even thought of at the time of static analysis. The implementation of this idea in a prototype system called the Shell for Interview Systems (SIS) is described. The interview systems generated with SIS (SIS-More, IIS-LD, and its improved version IIS-DB) are presented.<>