Abstract: Ontology mapping created by end-users contains class hierarchy structure knowledge of the domain，and using that knowledge can produce new domainal ontology mapping.Current system considers the complexity of class hierarchy and applies “equals to” relationship.But in the inference process，current system deletes inconsistent knowledge arbitrarily which would impact the recall.Aiming at solving those problems，this paper proposes a new concept of reliability of source.Besides，this paper re-designs inferences on class hierarchy knowledge base.The experiment result shows that revised system is more precise than the original one in knowledge structure，and can produce results with higher recall while maintaining high precision.

摘要： 基层用户产生的本体映射关系蕴含着相关领域的类层次结构，利用这部分知识可用于新映射关系的产生。现有的系统考虑了类层次关系的复杂性，加入了等价关系的应用，但是在知识库处理中的前向推理阶段，对矛盾知识仅是简单删除，从而影响到查全率。针对这些问题，提出了来源可信度以精确知识的可信程度，并重新设计了类层次知识库的矛盾知识推理规则。实验表明，改进的系统使知识结构更加精确，在保证查准率的同时可以明显改善查全率。