Cirrus: Inducing Subject Models from Protocol Data
Abstract
Cirrus is a system for aiding the analysis of protocol data. It assumes that subjects are following a non-deterministic procedure or plan, which is specified for it by the user. It uses machine learning techniques to find a deterministic version of the given procedure that maximizes the fit between the procedure's actions and the subject's actions, as recorded in the protocol given to Cirrus for analysis. This paper describes the theory of problem solving that Cirrus is based on, the current implementation, its analysis of several protocols, and our plans for future improvements. (kt)
Document Details
- Document Type
- Technical Report
- Publication Date
- Aug 16, 1988
- Accession Number
- ADA219148
Entities
People
- Bernadette Kowalski
- Kurt VanLehn
Organizations
- Carnegie Mellon University