Conventional query optimization usually minimizes the total time for complete evaluation of a query. We argue that this criterion may not be the only or even necessary requirement. In some interactive environments, a complete answer set may not be required and quick responses to queries can be very desirable. Minimizing the processing time for obtaining a single answer, a few answers, or most answers are identified as other potential optimization objectives. We attempt these optimization

more »

... s by adopting a strategy that tist looks at places where answers are most concentrated We are implementing a system (QAus) which provides quick answers to the user. The design of QAns is a loosely couple of a query transformer to a conventional query optimizer. The idea is based on semantic que~optimization. However, we introduce an additional logical equivalence in the query transformation process. We target the transformations towards reducing the cost of the set of more expensive operations in the original query plan. Inexact, as well as exaet rules are allowed in the transformation. The proliferation of inexact rules also leads to an issue of the "goodness" of such rules. We propose an informationtheoretic critetion for incorporating inexact rules in the system.