Query Rewriting Algorithms for Computing Credible Query Answers over Annotated Inconsistent Database

摘要

Managing and querying inconsistent database is a challenge problem: approaches of picking sure part or selecting one from the conflicting tuples result in information lose, while methods of computing all possible query answers can be meaningless because of the little probability of each possible query answer. We present an approach named Annotation Based Query Answer over Inconsistent Database which tries to calculate proper answer by distinguishing inconsistent data from consistent ones in the answer with annotations. It can correctly tell user inconsistency of query result down to attribute level when only functional dependency is considered. In this approach, information is preserved while query answer is one single. In this paper, we propose a method of query rewriting to compute Annotation Based Query Answer for any given SQL query without aggregation function and correlated sub query. Through the query rewriting, this approach doesn't require a new query language and can be easily embedded into existing database applications. Except for the information preserving, the experimental results both on TPC-H database and synthesized database show the effectiveness and applicability of our approach*