结合平移关系嵌入和CNN的知识图谱补全  

Knowledge Base Completion Based on Transitional Relation Embedding via CNN

作　　者：陈新元 谢晟祎[3] 陈庆强 刘羽 CHEN Xinyuan;XIE Shengyi;CHEN Qingqiang;LIU Yu(College of Computer and Control Engineering,Minjiang University,Fuzhou,Fujian 350100,China;Department of Information Engineering,Fuzhou Melbourne Polytechnic,Fuzhou,Fujian 350100,China;Experimental Training Center,Fujian Vocational College of Agriculturew Fuzhou,Fujian 35030,China;School of Information Science and Engineering,Fujian University of Technology,Fuzhou,Fujian 350100,China;Modern Education Technical Center,Fuzhou Melbourne Polytechnic,Fuzhou,Fujian 350100,China)

机构地区：[1]闽江学院计算机与控制工程学院,福建福州350100 [2]福州墨尔本理工职业学院信息工程系,福建福州350100 [3]福建农业职业技术学院实验实训中心,福建福州350300 [4]福建工程学院信息科学与工程学院,福建福州350100 [5]福州墨尔本理工职业学院现代教育技术中心,福建福州350100

出　　处：《中文信息学报》2021年第1期54-63,共10页Journal of Chinese Information Processing

基　　金：福建省中青年教师教育科研项目(JAT191663)。

摘　　要：为解决基于翻译机制的知识图谱补全模型在处理复杂关系时的性能局限,该文提出一种ATREC(algorithm based on transitional relation embedding via CNN)算法,将三元组的实体和关系映射至低维向量空间,并将不同的关系特征与头/尾实体融合,将原始三元组和融合三元组的嵌入表示合并为6列k维矩阵,使用卷积神经网络(CNN)降低参数规模,提取特征后拼接、赋权并评分。链路预测和三元组分类的实验结果表明,ATREC在较大规模数据集和复杂关系上相较主流算法有一定性能提升。To enhance the knowledge base(KB) completion regarding complex relations or nodes with high indegree or outdegree, an algorithm called ATREC(algorithm based on Transitional Relation Embedding via CNN) is proposed. In this method, entities and relations from triplets are first mapped into low-dimensional vector spaces. After relational fusion, then features from different relations are integrated into heads and tails, thus forming fused triplet representations. These representations of triplets are concatenated with original representations, forming a 6-column, k-dimensional matrices which serves as the input for convolution neural network(CNN). Experiments show that ATREC performs better than some state-of-the-art models especially when scaling up to relatively larger datasets and on relations with high cardinalities.

关 键 词：知识图谱补全 知识表示 CNN 翻译机制 链路预测 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]