Abstract: In view of the existing measurement method in the consideration not comprehensive and the calculation of index weights are determined on the basis of experience, this paper presents the geographic ontology concept semantic similarity measurement model based on PSO-BP—BP neural network optimized by particle swarm optimization. The model uses the properties of ontology, ontology structure and semantic relationship similarity, combines with the comprehensive weighted information calculation concept similarity. At the same time, the particle swarm optimization algorithm is used to optimize the BP neural network to obtain the factor weight, avoiding artificial subjective interference to determine factors weights in the existing methods. Finally, from the basic concepts of geographic information extracted 200 groups of samples, with 190 of group as the training set, the neural network model is trained to obtain the value of weights and the remaining 10 groups as a test set. Comparing the new model with several commonly used algorithms, by analyzing the correlation coefficient between algorithm results and expert judge results of the test set, it shows that the new model can more correctly solve the similarity of the concept of geographic ontology, in line with the characteristics of human cognition, more effective.

Key words: semantic similarity measurement, geographic ontology, Back Propagation（BP） neural network, Particle Swarm Optimization（PSO）

摘要： 针对现有度量方法中考虑因素不够全面和因子权重计算依据经验确定的不足，提出粒子群优化BP神经网络（PSO-BP）的地理本体概念语义相似度度量模型。该模型利用本体属性、本体结构和语义关系的相似度，结合权重信息计算概念的综合相似度；同时，利用粒子群算法优化的BP神经网络获取因子权重，避免现有方法中因子权重确定的人为主观干扰。最后，从基础地理信息概念中提取出200组样本，用其中190组作为训练集，对神经网络模型进行训练，以获取权重；剩余10组作为测试集。将该模型和几种常用算法进行对比，通过分析测试集的各算法求解结果和专家判定结果之间的相关系数，结果表明该模型计算地理本体概念的相似度更为准确，符合人类认知特性，效果更好。

关键词: 语义相似度度量, 地理本体, 反向传播（BP）神经网络, 粒子群算法