空间关系词汇与地理实体要素类型的语义约束关系构建方法

为克服依赖常识人工归纳总结的局限性,采用定性和定量相结合的方式,自动构建空间关系词汇与地理要素类型的语义约束关系。首先,以《中国大百科全书(地理版)》空间关系标注语料库为基础数据,采用Overlap语义相关度计量方法,挖掘空间关系词汇与地理要素类型约束关系的先验知识;然后以地理要素分类体系提供的概念语义相关性对先验知识进行扩展;最后采用本体工具Protégé建立相关知识库。

Semantic Relation Between Spatial Relation Terms and Feature Types of Geographical Entities

We propose a qualitative and quantitative approach to automatica construction of the semantic relation between spatial relation terms and feature types of geographical entities.Firstly,all spatial relations annotated in the Corpus are extracted as resource data,and then the “Overlap” similarity measure is used to define the relevance between spatial relation terms and geographical feature types.Secondly,the relevance is introduced as an original weight for the semantic distance computation of geographical concepts in the classification scheme of geographical feature types.Finally,a knowledge base is developed with protégé to formally represent and visualize the classification schemes,spatial relation terms and semantic relations with geographical feature types.This study indicates that our proposed approach can effectively obtain meaningful results.It is noted that the annotation quality of the corpus and the classification granularity of geographical entities have great effect on the performance.