Introduction - Repairing and Querying Databases under Aggregate Constraints - page 1

Chapter 1
Introduction
Abstract In this chapter, we make the reader acquainted with the problem of ex-
tracting reliable information from databases violating a given set of integrity con-
straints. After the presentation of the main issues related to repairing and querying
inconsistent databases, we introduce the specific problem of repairing and query-
ing numerical databases violating a particular class of integrity constraints (called
aggregate constraints ), whose investigation will be the main topic of the following
chapters. Finally, we summarize the related work in the literature dealing with the
problem of extracting reliable information from inconsistent databases, and provide
a description of the organization of the whole topic.
1.1 Dealing with Inconsistent Data
Integrity constraints are the mechanism employed in databases to guarantee that
available data correctly model the outside world. Databases may become inconsis-
tent w.r.t. the set of integrity constraints defined for several reasons. Commercial
database management systems avoid inconsistency by aborting updates or trans-
actions yielding to an integrity constraint violation. However, often integrity con-
straints satisfaction cannot be guaranteed by employing this approach. For instance,
when knowledge frommultiple sources is integrated, as in the contexts of data ware-
housing, database integration and automated reasoning systems, it is not possible to
guarantee the consistency on the integrated information applying the traditional ap-
proach. More in detail, when different source databases are integrated, although ev-
ery source database is consistent with respect to a given set of integrity constraints,
in the resulting integrated database many different kinds of discrepancies may arise.
In particular, possible discrepancies are due to (i) different sets of integrity con-
straints that are satisfied by different sources, and (ii) constraints that may be glob-
ally violated, even if every source database locally satisfies the same integrity con-
straints. In this case, there is no update that can be rejected in order to guarantee the
consistency of the integrated database as this database instance is not resulting from