QUALITATIVE QUERYING OF PHYSICAL PROCESS SIMULATIONS

The preliminary design of a query system for the computer simulation o f physical processes is presented. Most engineering simulations produc e vast amounts of quantitative data. It often requires special trainin g to interpret the data and many of them may not be directly relevant to a particular decision. Human beings usually find qualitative explan ations easier to absorb than large quantities of detail. The query sys tem described in the paper is a preliminary to the production of an in terrogation system which can provide both qualitative answers and expl anations. In the first part of the system, natural language queries ar e transformed into standard query types using a neural net. In the sec ond part qualitative replies are formed from neural nets which are tra ined by data from the simulation. An object oriented hierarchy which f acilitates the mappings of quantitative data into qualitative attribut es is presented. The development of a more detailed and sophisticated explanation system is also discussed.