Coping with Data Inconsistencies in the Integration of Heterogenous Data Sources
This research examines the problem of inconsistent data when integrating information from multiple sources into a unified view. Data inconsistencies undermine the ability to provide meaningful query responses based on the integrated data. The study reviews current techniques for handling inconsistent data including domain-specific data cleaning and declarative methods that provide answers despite integrity violations. A key challenge identified is modeling data consistency and ensuring clean integrated data. Data integration systems based on a global schema must carefully map heterogeneous sources to that schema. However, dependencies in the integrated data can prevent attaining consistency due to issues like conflicting facts from different sources. The research summarizes various proposed approaches for resolving inconsistencies through data cleaning, integrity constraints, and dependency mapping techniques. However, outstanding challenges remain regarding accuracy, availability, timeliness, and other data quality restrictions of autonomous sources. Additional research is needed to develop more automated ways of reconciling inconsistencies from source data with the requirements of the global schema. The ability to provide high-quality integrated data is crucial for organizations to maximize the value of their information assets. This research aims to promote further investigation into semi-automated remediation of inconsistencies and leveraging source data quality metrics to aid the integration process. Overcoming inconsistencies is critical to enabling unified views and meaningful analytics from merged cross-organizational data.
