Cooperative Query Answering with Density Scores
Felix Naumann, Ulf Leser · edoc Publication server (Humboldt University of Berlin) · 2000
Mediator-based information systems answer global queries by rewriting them into a combination of queries against physical data sources. One assumption in most systems is that only such combinations are considered as valid that obtain values for each selected attribute of the query. Another assumption is that systems must compute and execute all valid combinations, i.e., they strive to retrieve all possible answers. These assumptions frequently lead to user frustration: First, in many scenarios an incomplete answer is much more appreciated than no answer at all. Second, if many valid combinations exist, it is very time-consuming to execute them all. We present a cooperative query planning method that avoids both problems. First, it treats incomplete and complete source combinations in a logically equivalent manner, i.e., incomplete answers are also considered. Second, it ranks and selects combinations based on their information density, i.e., based on the expected usefulness of the results.