Data Conflict Resolution Using Trust Mappings

TL;DR

This paper introduces a novel, efficient method for resolving data conflicts in community databases using trust mappings, ensuring globally consistent snapshots through a PTIME algorithm based on logic programming semantics.

Contribution

It presents the first PTIME algorithm for conflict resolution in community databases, leveraging stable model semantics for a principled approach.

Findings

The proposed algorithm guarantees a globally consistent snapshot.

Conflict resolution is achieved in polynomial time.

Extensions to negative beliefs are discussed, with some proven hard.

Abstract

In massively collaborative projects such as scientific or community databases, users often need to agree or disagree on the content of individual data items. On the other hand, trust relationships often exist between users, allowing them to accept or reject other users' beliefs by default. As those trust relationships become complex, however, it becomes difficult to define and compute a consistent snapshot of the conflicting information. Previous solutions to a related problem, the update reconciliation problem, are dependent on the order in which the updates are processed and, therefore, do not guarantee a globally consistent snapshot. This paper proposes the first principled solution to the automatic conflict resolution problem in a community database. Our semantics is based on the certain tuples of all stable models of a logic program. While evaluating stable models in general is well known to be hard, even for very simple logic programs, we show that the conflict resolution problem admits a PTIME solution. To the best of our knowledge, ours is the first PTIME algorithm that allows conflict resolution in a principled way. We further discuss extensions to negative beliefs and prove that some of these extensions are hard. This work is done in the context of the BeliefDB project at the University of Washington, which focuses on the efficient management of conflicts in community databases.