Abstract: Aiming at the problem of " knowledge island " between water conservancy project data texts, a dam safety knowledge graph construction method based on UIE framework and SE-RE-Joint model is proposed. The seven-step method is used to construct a multi-layer domain ontology library of main structure, affiliated buildings and comprehensive evaluation. After data preprocessing, the preliminary entity recognition is realized by fine-tuning of small samples. Through semantic enhanced coding, relying on bidirectional GRU to capture sequence dependence and CRF to optimize label sequence consistency, entity recognition and relationship extraction are completed. Finally, Neo4j graph database is used to store knowledge. The experimental results show that UIE+SE-RE-Joint extracts 25,000 entities and 31,000 pairs of relationships, and the prediction accuracy of the model exceeds 84%. The four types of errors of entity conflict, boundary blur, type confusion and relationship misjudgment are significantly improved compared with the comparison model. The map can support knowledge visualization and data retrieval, and complete knowledge dynamic updating, knowledge reasoning and risk positioning through Python language, which can assist the intelligent diagnosis of dam safety.