|
Edward
Brent, Albert F. Anderson, and Pawel Slusarz For inexperienced users of complex data sets, mastering the metadata can be a formidable task. What variables are available in the data? Are those variables available for the year of interest? How is this concept measured? How can discrepancies in measures for different years be resolved? This paper describes one component of an integrated approach in which intelligent agents permit the user to issue broad queries delegating the details to the agent; case-based reasoning guides the user to relevant examples; machine learning permits successful queries to be added to the program's expanding knowledge base for help with future queries; and expert systems provide advice to the user on a range of issues. Often the easiest way to help a user create a new query is to show them a similar query they can use with only minor modifications. In this paper we describe and illustrate a prototype module using case-based reasoning to identify similar queries from a database of previous queries This module is illustrated for the PDQ-Explore system for providing rapid intelligent access to the IPUMS (Integrated Public Use Microdata Series) dataset. Edward Brent Albert F. Anderson Pawel Slusarz |