Acta Scientiarum Naturalium Universitatis Pekinensis ›› 2023, Vol. 59 ›› Issue (3): 434-444.DOI: 10.13209/j.0479-8023.2023.002

Previous Articles     Next Articles

A Spatially Constraint Negative Sample Generation Method for Geographic Knowledge Graph Embedding

GAO Yong, MENG Haohan, YE Chao   

  1. Institute of Remote Sensing and Geographic Information System, School of Earth and Space Sciences, Peking University, Beijing 100871
  • Received:2022-04-28 Revised:2022-05-11 Online:2023-05-20 Published:2023-05-20
  • Contact: GAO Yong, E-mail: gaoyong(at)pku.edu.cn

基于空间约束的地理知识图谱嵌入表示的负样本生成方法

高勇, 孟浩瀚, 叶超   

  1. 北京大学地球与空间科学学院遥感与地理信息系统研究所, 北京 100871
  • 通讯作者: 高勇, E-mail: gaoyong(at)pku.edu.cn
  • 基金资助:
    国家自然科学基金(41971331)资助

Abstract:

Geographic knowledge graph representation learning requires generating the corresponding negative samples based on the positive ones. However, traditional negative sample generation algorithms suffer from high error rate and poor adaption to geographic knowledge graph. Aimming at this problem, a spatially constraint negative sample generation method was proposed by modifying the modeling of spatial relations. Then the method was applied to different knowledge graph representation learning models to explore its suitability in geographic knowledge graph embedding. Results show that the proposed method has a low error rate and is suitable for two common types of knowledge graph representation models. The spatially constraint negative sample generation method will improve the accuracy of geographic knowledge graph representation learning, which helps to advance geographical research.

Key words: geographic knowledge graph, representation learning, spatial constraint, spatial relationship, place

摘要:

地理知识图谱的表示学习需要根据正样本生成对应的负样本, 然而传统的负样本生成算法存在错误率高、地理知识图谱适配性差的问题。针对这一问题, 调整空间关系在地理知识图谱中的表达方式, 提出基于空间约束的负样本生成方法, 并将该方法应用至不同的知识图谱表示学习模型, 探讨其在地理知识图谱表示学习中的适配性。结果表明, 该算法具有较低的错误率, 同时适用于常见的两类知识图谱表示模型, 能够提高地理知识图谱表示学习的精度, 有助于地理知识图谱在地理研究中发挥更重要的作用。

关键词: 地理知识图谱, 表示学习, 空间约束, 空间关系, 场所