For more information contact Luis Rocha at rocha@lanl.gov
Last Modified: May 15, 2000
An important issue in developing human-oriented interfaces has been to build querying models that allow non-expert users to query databases in a natural way. However, natural language includes many kinds of imprecise statements which are not compatible with the strict structure of relational databases. This talk presents an intelligent human-oriented interface based on fuzzy logic. The queries reasoning process uses four fuzzy translation rules of the meaning representation language PRUF (Possibilistic Relational Universal Fuzzy), as proposed by Zadeh.
A database with the 500 biggest non-financial Portuguese companies is used to obtain specific queries about business characteristics and to discuss the syntax and semantic concepts involved in our querying model. Illustrative examples of the four natural language queries, such as “is IBM very profitable?” or “are most private companies productive?”, are presented to show the capabilities of this human-oriented interface.