Repair diversification: A new approach for data repairing

In practice, data are often found to violate given integrity constraints, e.g., functional dependencies, and are hence inconsistent. To resolve such violations, data are to be restored to a consistent state, known as “repair”, while the number of possible repairs may be exponential. Previous works either consider optimal repair computation, to find one single repair that is (nearly) optimal w.r.t. some cost models, or discuss repair sampling, to randomly generate a repair from the space of all possible repairs.This paper makes a first effort to investigate repair diversification problem, which aims at generating a set of repairs by minimizing their costs and maximizing their diversity. There are several motivating scenarios where diversifying repairs is desirable. For example, in the recently proposed interactive repairing approach, repair diversification techniques can be employed to generate several representative repairs that are likely to occur (small cost), and at the same time, that are dissimilar to each other (high diversity). Repair diversification significantly differs from optimal repair computing and repair sampling in its framework and techniques. (1) Based on two natural diversification objectives, we formulate two versions of repair diversification problem, both modeled as bi-criteria optimization problem, and prove the complexity of their related decision problems. (2) We develop algorithms for diversification problems. These algorithms embed repair computation into the framework of diversification, and hence find desirable repairs without searching the whole repair space. (3) We conduct extensive performance studies, to verify the effectiveness and efficiency of our algorithms.