ISSN 1000-1239 CN 11-1777/TP

计算机研究与发展 ›› 2019, Vol. 56 ›› Issue (12): 2549-2561.doi: 10.7544/issn1000-1239.2019.20190648

所属专题: 2019大数据知识工程及应用专题

• 其他应用技术 • 上一篇    下一篇



  1. 1(内蒙古大学计算机学院 呼和浩特 010021);2(北方工业大学信息学院 北京 100144);3(中国科学技术信息研究所 北京 100038) (
  • 出版日期: 2019-12-01
  • 基金资助: 

Open Knowledge Graph Representation Learning Based on Neighbors and Semantic Affinity

Du Zhijuan1, Du Zhirong2, Wang Lu3   

  1. 1(College of Computer Science, Inner Mongolia University, Hohhot 010021);2(College of Information, North China University of Technology, Beijing 100144);3(Institute of Scientific and Technical Information of China, Beijing 100038)
  • Online: 2019-12-01

摘要: 知识图谱(knowledge graph, KG)打破了不同场景下的数据隔离,为实际应用提供基础支持.表示学习将KG转换到低维向量空间来为KG应用提供便利.然而,KG的表示学习目前存在2个问题:1)假设KG满足闭合世界假设,要求所有实体在训练中可见.实际上,大多数KG都在快速增长,例如DBPedia平均每天产生200个新实体.2)采用矩阵映射、卷积等复杂的语义交互方式提高模型的准确性,这样做也限制了模型的可扩展性.为此,针对允许新实体存在的开放KG,提出一种表示学习方法TransNS.它选取相关的邻居作为实体的属性来推断新实体,并在学习阶段利用实体之间的语义亲和力选择负例三元组来增强语义交互能力.5个传统数据集和8个新数据集对比了TransNS与最经典的表示学习方法,结果表明:TransNS在开放KG上表现良好,甚至在基准闭合KG上优于现有模型.

关键词: 知识图谱, 表示学习, 开放世界假设, 邻居, 语义亲和力

Abstract: Knowledge graph (KG) breaks the data isolation in different scenarios and provides basic support for the practical application. The representation learning transforms KG into the low-dimensional vector space to facilitate KG application. However, there are two problems in KG representation learning: 1)It is assumed that KG satisfies the closed-world assumption. It requires all entities to be visible during the training. In reality, most KGs are growing rapidly, e.g. a rate of 200 new entities per day in the DBPedia. 2)Complex semantic interaction, such as matrix projection and convolution, are used to improve the accuracy of the model and limit the scalability of the model. To this end, we propose a representation learning method TransNS for open KG that allows new entities to exist. It selects the related neighbors as the attribute of the entity to infer the new entity, and uses the semantic affinity between the entities to select the negative triple in the learning phase to enhance the semantic interaction capability. We compare our TransNS with the state-of-the-art baselines on 5 traditional and 8 new datasets. The results show that our TransNS performs well in the open KGs and even outperforms existing models on the baseline closed KGs.

Key words: knowledge graph, representation learning, open-world assumption, neighbors, semantic affinity