Nowadays, organization faces the problem of acquiring correct information from data asset to support decision-making. To solve this problem, a semantic correctness oriented relational data access method, namely SCORDA, is proposed. Firstly, ontology was used to describe the concept view, which is used in data access task, and description logic makes a suitable tradeoff between expressiveness and complexity. Secondly, a LAV+GAV mapping formalism was adopted to solve the schema heterogeneity and impedance mismatch problem. Furthermore, a special structure named dynamic ABox is provided to deal with the instance checking problem, and to ensure the semantic correctness of answers based on the model theory semantic of ontology. Finally, the feasibility of SCORDA method, through implementing a prototype system， was validated.