Abstract: Over the past few years, ontology-based semantic similarity has been widely used in many fields of information science. As such, methods for the calculation of semantic similarities from ontology have been receiving more and more attention. This paper analyzes principles, advantages, and disadvantages of existing methods for calculating semantic similarities, and it proposes a new method that assesses semantic similarity based on two properties of the Directed Acyclic Graph（DAG） structure of an ontology：the degree of overlap in paths from the root to the nodes corresponding to the given concepts and the depth of the Lowest Common Ancestor（LCA） node of these nodes. The proposed method is flexible and can be applied to ontologies of a wide variety of domains. It applies the proposed method to Gene Ontology（GO） and Plant Ontology（PO）, and it compares the results with those of two leading methods. Results show the correctness and effectiveness of the proposed method.

Key words: semantic similarity, ontology, directed acyclic graph

摘要： 基于本体的概念语义相似度近年来在信息科学的多个领域获得了广泛的应用，其计算方法也为诸多学者所关注。分析现有基于本体的概念语义相似度计算方法的工作原理和优缺点，提出一种对概念共享路径的重合度和概念最低共同祖先节点的深度进行综合加权的概念语义相似度算法。该算法灵活简便、可扩展性强，能够应用于不同类型的本体。使用基因本体和植物本体的部分数据进行了实验并与两种现有算法进行了比较，实验结果证明了提出的计算方法的正确性和有效性。

关键词: 语义相似度, 本体, 有向无环图