How would you approach a dataset that has conflicting information, and resolve discrepancies between sources?