Authors: Eric K. Jones, Aaron Roydhouse
Source: GZipped PostScript (262kb); Adobe PDF (249kb)
Large repositories of historical data about the natural world present novel opportunities to assist decision makers and problem solvers. Past cases or scenarios that are similar to a problem of current interest can often provide valuable insights or suggest solutions to the problem.
We are constructing a case retrieval system that is intended to give meteorologists fast access to relevant past weather situations. The system is to serve as a ``memory amplifier,'' allowing users to rapidly locate historical situations of interest. In particular, forecasters should be able to use the system to quickly retrieve situations that are similar to the current one in meteorologically significant respects, providing them with an additional source of information to supplement the output of numerical models. The case retriever is part of Metvuw Workbench, a system for intelligent retrieval and display of historical meteorological data.
In designing the Metvuw case retriever, we have addressed challenging problems in three areas: representation of spatial and temporal knowledge, case selection, and similarity assessment. We have also identified a number of lessons for the design of intelligent databases in natural resource domains.