The reasoning performed to verify the correctness of the RDFS ontologies is the task with high computational cost. The task will become more complex in the scenario of existence of additional constraints. This paper presents an approach that can extract the RDFS schema without changing the reasoning results. This approach is based on the analysis of the dependency relationships between the constraints. To capture the precise semantics of the RDFS schema, firstly, the schema elements and the constraints are formalized into first-order formulas expressed as disjunctive embedded dependencies and the constraint depandency graph is established according to the interaction between the constraints. Then, the strategies for deleting the edges and the nodes that are irrelevant to the reasoning tasks are applied. Finally, the RDFS sub-schema is obtained through the reconstruction process. The proposed approach enables the reasoning to be carried out on the extracted small-scale ontologies. The experiment results show that the proposed approach can significantly improve the efficiency of the RDFS ontology validation. Compared to the reasoning time, the average time (0.60s) consumed by the extraction process is almost negligible, while the efficiency increasement ranges from 2.00 times to 22.97 times.