Augmented transition networks as a representation for knowledge-based history-taking systems.

Numerous history-taking systems have been built to automate the medical history-taking process. These systems differ in their control methods, input and output modalities, and kinds of questions asked. Thus, there has emerged no standard way of representing interviewing knowledge--the expert knowledge used to govern the sequence of questions asked in an interview. This paper discusses how we use an augmented transition network (ATN) to represent the knowledge of a speech-driven automated history-taking program, Q-MED, and how, more generally, ATNs could be used as a representation for any knowledge-based history-taking system. We identify three characteristics of ATNs that facilitate the use of ATNs in interviewing systems: explicitness, hierarchical structure, and generality.