In this paper, we propose an approach to DATA fusion to enhance the accuracy of DATA INTEGRATION. The proposed approach uses the information in the relationships between entities to find more evidence for the correctness or incorrectness of the values generated by different DATA sources. We also define some concepts and investigate the different methods for identifying relationships between entities. Then, we consider how to use these relationships to increase the accuracy of the conflict resolution process. Unlike many existing approaches, our proposed approach is at the high level of DATA abstraction. Using the information there exists at the high levels of DATA abstraction allows us to provide sufficient evidences where DATA is incomplete and there is no reliable source for the particular object. The evaluation results show that our proposed approach outperforms existing conflict resolution techniques.