QUALITATIVE QUERYING OF PHYSICAL PROCESS SIMULATIONS

The preliminary design of a query system for the computer simulation of physical processes is presented. Most engineering simulations produce vast amounts of quantitative data. It often requires special training to interpret the data and many of them may not be directly relevant to a particular decision. Human beings usually find qualitative explanations easier to absorb than large quantities of detail. The query system described in the paper is a preliminary to the production of an interrogation system which can provide both qualitative answers and explanations. In the first part of the system, natural language queries are transformed into standard query types using a neural net. In the second part qualitative replies are formed from neural nets which are trained by data from the simulation. An object oriented hierarchy which facilitates the mappings of quantitative data into qualitative attributes is presented. The development of a more detailed and sophisticated explanation system is also discussed.