Trans系列知识表示方法发展研究  

Research on Development of TransX Knowledge Representation Methods

作　　者：杜晓娟 陶以政[1] 李龚亮 DU Xiaojuan;TAO Yizheng;LI Gongliang(Institute of Computer Application,China Academy of Engineering Physics,Mianyang 621999)

机构地区：[1]中国工程物理研究院计算机应用研究所,绵阳621999

出　　处：《计算机与数字工程》2023年第11期2515-2517,共3页Computer & Digital Engineering

基　　金：国家重点研发计划(编号:2016YFB0801001)资助。

摘　　要：近年来,知识图谱被普遍应用于搜索引擎和智能推荐系统,而知识表示作为知识图谱关键技术得到了广泛的关注。关于知识表示方法的研究也一直在进行,其中以TransE为代表的基于翻译模型的知识表示方法有着容易理解、复杂度低、计算效率高等优点,而且与此同时,具有较好的知识表达能力。文章对Trans系列模型进行了总结,着重分析了不同模型之间的改进思路并对其知识表示能力进行了对比,为进一步的改进提供参考。In recent years,knowledge graphs have been widely used in search engines and intelligent recommendation sys⁃tems,and knowledge representation as a key technology of knowledge graphs has received extensive attention.The knowledge repre⁃sentation method based on translation models represented by TransE has the advantages of easy understanding,low complexity,and high computational efficiency.At the same time,TransX has good knowledge expression preformance.This paper summarizes the TransX models,and analyzes the improvement between different models and compares their knowledge representation capabilities.It is a reference for the advancement of knowledge representation methods.

关 键 词：知识图谱 知识表示 TransE TransH TransR 

分 类 号：O141.4[理学—数学]