地理要素类别语义相似度动态加权计算方法

A dynamic weighted model for semantic similarity measurement between geographic feature categories

语义相似度计算是解决地理要素类别语义异质问题的关键技术,在地理数据共享与交换应用中起着重要的作用。本文面向基础地理信息领域,针对相同特征属性在不同地理要素类别语义中存在重要性差异的特点,提出了一种基于动态权重的地理要素类别语义相似度算法。本文算法引入词频-逆向文件频率,利用属性值的特殊性,计算特征属性的动态权重,根据不同类型的特征属性提出相应的相似度算法,重点分析多值复杂型属性值的分解及相似度算法,得到地理要素类别间的相似度。最后,从基础地理要素类别中选择200组样本对计算语义相似度,并与其他4种相似度算法进行对比分析。试验结果表明,本文算法能够有效地反映特征属性的重要性差异,获得更准确合理的地理要素类别的语义相似度。

Semantic similarity is a key technology to solve the problem of semantic heterogeneity of geographic feature categories,and plays an important role in geographic data sharing and exchange applications.In this article,a semantic similarity calculation model of geographic feature categories based on dynamic weights is proposed to represent the difference in importance of semantic properties among geographic feature categories for the need of fundamental geographical domain application.TF-IDF algorithm is introduced and the particularity of the property value is used to calculate the dynamic weight of a semantic property.And the similarity model between a pair of complex properties is proposed,and then the final similarity between geographic feature categories is calculated.Finally,200 pairs of samples are selected from the basic geographic element categories to calculate the semantic similarity,and compared with other similarity calculation models.The experimental results show that the model proposed in this article can effectively reflect the importance difference of semantic properties and obtain more reasonable semantic similarity between geographic feature categories.