[Paper Review] Principles of Progress Indicators for Database Repairing.

Ester Livshits, Ihab F. Ilyas+2 more|arXiv (Cornell University)|Apr 13, 2019

Data Quality and ManagementDecision Sciences28 references3 citations

TL;DR

This paper proposes rational, tractable inconsistency measures for database repairing by formalizing postulates that govern how inconsistency should decrease during cleaning. It introduces a measure based on tuple deletions under denial constraints that ensures steady, continuous progress with computational efficiency, enabling practical online progress indicators.

ABSTRACT

How should a cleaning system measure the amount of inconsistency in the database? Proper measures are important for quantifying the progress made in the cleaning process relative to the remaining effort and resources required. Similarly to effective progress bars in interactive systems, inconsistency should ideally diminish steadily and continuously while aiming to avoid jitters and jumps. Moreover, measures should be computationally tractable towards online applicability. Building on past research on inconsistency measures for knowledge bases, we embark on a systematic investigation of the rationality postulates of inconsistency measures in the context of database repair systems. The vision is to establish postulates that take into consideration the interplay between the database tuples, the integrity constraints, and the space of atomic repairing operations, and that guide the design of practical measures and progress indicators. Toward that, we shed light on classic measures by examining them against a set of rationality postulates. We also demonstrate the feasibility of the postulates by proposing a measure that is both rational and tractable for the general class of denial constraints, under the restriction of the repairing operations to tuple deletions.

Motivation & Objective

• To define rationality postulates for inconsistency measures in database repairing that account for tuples, integrity constraints, and atomic repair operations.
• To evaluate existing inconsistency measures against these postulates to assess their rationality in the context of database cleaning.
• To design a tractable inconsistency measure that ensures continuous, jitter-free progress during database repair.
• To demonstrate feasibility of the proposed framework on denial constraints with tuple deletion as the only repair operation.

Proposed method

• Formalize a set of rationality postulates that inconsistency measures should satisfy, considering the interplay between tuples, integrity constraints, and repair operations.
• Adapt and analyze classic inconsistency measures from knowledge base research in the context of database repairing.
• Propose a new inconsistency measure based on the number of tuples involved in violations, using only tuple deletions as repair operations.
• Prove that the proposed measure is both rational (satisfies all postulates) and computationally tractable for denial constraints.
• Demonstrate that the measure ensures monotonic and continuous decrease in inconsistency during repair, avoiding jumps or oscillations.

Experimental results

Research questions

• RQ1 Which postulates should an inconsistency measure satisfy to be considered rational in the context of database repairing?
• RQ2 How do existing inconsistency measures perform when evaluated against the proposed rationality postulates?
• RQ3 Can a tractable inconsistency measure be designed that ensures steady, continuous progress during database repair?
• RQ4 Is it feasible to construct such a measure under the restriction of tuple deletions and denial constraints?

Key findings

• The proposed inconsistency measure satisfies all defined rationality postulates, ensuring logical consistency in progress tracking.
• The measure is computationally tractable for the general class of denial constraints when only tuple deletions are allowed as repair operations.
• The inconsistency level decreases monotonically and continuously during the repair process, avoiding jitters or jumps.
• The framework enables the design of practical, online progress indicators for database cleaning systems.