A Distributed Technique for Repairing Inconsistencies of Functional Dependencies

Dependencies play an important role in database and have been used in schema design, data cleaning and integration, among others. One of the fundamental problems in data quality is repairing inconsistencies for given dependencies. Given an instance I of a relational schema R and a set Σ of dependencies rules, we want to repair all cells in the instance I that violate rules in Σ, such that all rules in Σ are satisfied. It is typically difficult to do so efficiently since rules may be closely relate to each other. This paper presents methods for repairing violations of functional dependencies (FDs) in distributed data with techniques of equivalent classes. When data in I are distributed, repairing inconsistencies often necessarily requires shipping original data from one site to another. We group the data by rolling up the relational attributes in order to reduce the amount of data transmission. We also present an algorithm to handle equivalent classes for multiple FDs with the same right hand side attributes.