Efficient query repair for aggregate constraints

Abstract

In many real-world scenarios, query results must satisfy domain-specific constraints such as fairness or financial stability. For example, selecting interview candidates based on their qualifications may require that at least a given percentage be female, or a report on purchase costs may need to ensure the average cost stays below a liability threshold. These requirements can be expressed as constraints over an arithmetic combination of aggregates evaluated on the result of the query.

This thesis studies how to repair a query to fulfill such constraints by modifying the filter predicates of the query. These constraints are non monotone and more complex than those considered in prior work, such as query based explanations for missing answers or fairness enforcement in query results. The constraints considered in this thesis invalidate many existing optimizations considered in prior work. The work in this thesis introduces a novel query repair technique that computes the top-k candidate repairs with respect to their distance to the user query. These techniques leverage materialization and data clustering to avoid unnecessary computation. It also exploits bounds on sets of candidate solutions and interval arithmetic to efficiently prune the search space. Experimental evaluation on real-world and benchmark datasets shows that the proposed pruning technique significantly outperforms baselines that consider a single candidate at a time.

More information

Identifiers

ORCID for Shatha Saad Algarni: orcid.org/0009-0004-9831-6034

ORCID for Age Chapman: orcid.org/0000-0002-3814-2587

ORCID for Steffen Staab: orcid.org/0000-0002-0780-4154

Catalogue record

Export record