动态分块调节机制下的大规模解剖学本体匹配

传统的匹配系统采用固定分块的方式处理大规模解剖学本体，普遍存在语义信息的丢失，影响了匹配效果。为此，提出一种动态分块调节机制，将按匹配情况确定的实体不断地重新分配到目标块中，动态地调节各个分块，从而尽可能地保留语义完整性。此外，针对该问题计算复杂度高的特点，引入了紧凑进化算法对子匹配任务中的阈值以及进行实体重新分配的块标志位优化，并设计了一种精英解参与的概率向量更新方式对该算法进行改进。实验在OAEI(ontology alignment evaluation initiative)的Anatomy测试集上进行，验证了所提方法对匹配结果质量的提升。此外，和其他匹配系统的对比也展示了所构建匹配系统的先进性。

Large scale anatomical ontology matching under dynamic partition adjustment

Traditional matching systems adopt fixed partitioning methods, which easily lead to the lost of semantic information, to match the large scale anatomical ontologies. In order to preserve the semantic information as completely as possible and improve the quality of matching results, this paper developed a dynamic partition adjustment mechanism. It adjusted the divided blocks dynamically by continuously redistributing the determined entities to the target blocks. In view of the high computation complexity of this matching model, this paper proposed an improved compact evolutionary algorithm to optimize the thresholds of sub-matching tasks and the block flags for redistribution. Experiments on the Anatomy test set of OAEI verify that the dynamic partition adjustment mechanism and improved compact evolutionary algorithm can enhance the alignment quality of matching system. Furthermore, a comparison with other matching systems demonstrates the advance of the proposed system.