• 北大核心期刊(《中文核心期刊要目总览》2017版)
  • 中国科技核心期刊(中国科技论文统计源期刊)
  • JST 日本科学技术振兴机构数据库(日)收录期刊

留言板

尊敬的读者、作者、审稿人, 关于本刊的投稿、审稿、编辑和出版的任何问题, 您可以本页添加留言。我们将尽快给您答复。谢谢您的支持!

姓名
邮箱
手机号码
标题
留言内容
验证码

基于表示学习的多关系型知识图谱推理算法

袁培 王儒敬

袁培, 王儒敬. 基于表示学习的多关系型知识图谱推理算法[J]. 微电子学与计算机, 2022, 39(4): 75-82. doi: 10.19304/J.ISSN1000-7180.2021.1089
引用本文: 袁培, 王儒敬. 基于表示学习的多关系型知识图谱推理算法[J]. 微电子学与计算机, 2022, 39(4): 75-82. doi: 10.19304/J.ISSN1000-7180.2021.1089
YUAN Pei, WANG Rujing. Multi-relational knowledge graph reasoning algorithm based on representation learning[J]. Microelectronics & Computer, 2022, 39(4): 75-82. doi: 10.19304/J.ISSN1000-7180.2021.1089
Citation: YUAN Pei, WANG Rujing. Multi-relational knowledge graph reasoning algorithm based on representation learning[J]. Microelectronics & Computer, 2022, 39(4): 75-82. doi: 10.19304/J.ISSN1000-7180.2021.1089

基于表示学习的多关系型知识图谱推理算法

doi: 10.19304/J.ISSN1000-7180.2021.1089
基金项目: 

国家重点研发计划项目 2019YFE0125700

详细信息
    作者简介:

    袁培  女,(1996-),硕士研究生.研究方向为知识图谱

    通讯作者:

    王儒敬(通讯作者)  男,(1964-),博士,研究员.研究方向为智能农业.E-mail: rjwang@ilm.ac.cn

  • 中图分类号: TP391

Multi-relational knowledge graph reasoning algorithm based on representation learning

  • 摘要:

    目前知识图谱的推理方法中,表示学习尤其是基于翻译的TransE系列算法取得了优异表现.其相关论文大都关注实体推理,然而关系推理作为知识图谱补全的关键技术值得受到关注与研究.与此同时,在规模不断扩大、知识来源更加多样化的知识图谱中,关系种类繁多且类型复杂,单个关系在全体三元组中的出现频率进一步降低,这为关系推理增加了难度.因此针对多关系型知识图谱,基于TransE模型并侧重知识图谱三元组中关系的推理,提出一种新的关系建模方法,通过调整向量空间中实体向量与关系向量间的组织结构,缓解多映射属性关系中不同种类的关系争抢同一向量的问题.然后又与其他方法结合,使新的模型在实体推理上具备可行性.通过在公开的FB15k数据集以及自行从网络中抽取得到的中文数据集上展开的知识推理实验,从关系推理准确率与实体推理准确率等指标与相似方法进行对比,均取得了良好的表现,成功验证了算法的有效性与先进性.

     

  • 图 1  TransE模型的核心思想

    Figure 1.  Main idea of the TransE model

    图 2  TransRe模型的核心思想

    Figure 2.  Main idea of the TransRe model

    图 3  TransRe模型中尾实体的推理

    Figure 3.  Tail entity reasoning of the TransRe model

    表  1  FB15k数据集数据量

    Table  1.   The statistics of the FB15k dataset

    实体数量 关系数量 三元组数量
    训练集 测试集
    14 951 1 345 483 142 59 071
    下载: 导出CSV

    表  2  中文数据集数据量

    Table  2.   The statistics of the Chinese dataset

    实体数量 关系数量 三元组数量
    训练集 测试集
    22 754 559 99 438 25 926
    下载: 导出CSV

    表  3  数据集关系数据量对比

    Table  3.   The relations′statistics of the corpora

    关系数量 三元组数量 平均出现次数
    FB15k 1 345 542 213 403
    中文数据集 559 125 364 224
    FB15k-237 237 310 116 1 309
    NELL-995 200 154 213 771
    下载: 导出CSV

    表  4  知识推理实验环境

    Table  4.   Experimental environment

    操作系统 Windows10
    CPU Intel(R) Core(TM) i7-10710U CPU @ 1.10GHz 1.61 GHz
    IDE PyCharm
    Python 3.8
    Numpy 1.20.2
    内存 16GB
    下载: 导出CSV

    表  5  各模型所用参数

    Table  5.   Parameters used by models

    数据集 模型 B k γ α
    FB15k TransRe 960 150 0.2 0.005
    TransReF 960 200 0.5 0.0025
    中文数据集 TransE 960 50 1 0.01
    TransF 960 100 0.7 0.002 5
    TransRe 960 150 0.5 0.005
    TransReF 960 200 0.7 0.005
    下载: 导出CSV

    表  6  关系推理的结果

    Table  6.   Experiment results on relational reasoning

    hits@1/% hits@10/% MR MRR
    TransE 43.3 67.3 379 0.543
    TransF 51.9 84.6 80 0.644
    IWM[14] 41.3 80.3 * 0.555
    UGRW[15] 67.4 69.3 * 0.603
    TDSR[15] 64.8 76.3 * 0.529
    TransRe 64.1 90.1 95 0.744
    TransReF 55.5 85.7 79 0.668
    下载: 导出CSV

    表  7  TransRe模型在不同类型关系上关系推理的结果

    Table  7.   Detailedexperiment results on relational reasoning of the TransRe model

    1∶1 1∶M M∶1 M∶M
    hits@1/% 23.7 34.4 39.7 37.2
    hits@10/% 88.4 98.0 98.4 74.8
    下载: 导出CSV

    表  8  关系推理的结果

    Table  8.   Experiment results on relational reasoning

    hits@1/% hits@10/% MR
    TransE 38.8 67.0 44
    TransF 51.0 76.5 37
    TransRe 70.4 89.8 27
    TransReF 54.4 77.0 35
    下载: 导出CSV

    表  9  实体推理的结果

    Table  9.   Experiment results on entity reasoning

    hits@10/% MR
    TransE[5] 34.9 243
    TransH[6] 42.5 211
    CTransR[7] 44.0 233
    TransF[8] 40.5 220
    TransD[9] 49.4 211
    TranSparse[10] 50.3 216
    TransRe 34.5 234
    TransReF 48.5 203
    下载: 导出CSV
  • [1] BENGIO Y, COURVILLE A, VINCENT P. Representation learning: A review and new perspectives[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2013, 35(8): 1798-1828. DOI: 10.1109/TPAMI.2013.50.
    [2] 董永峰, 刘超, 王利琴, 等. 融合多跳关系路径信息的关系推理方法[J]. 计算机应用, 2021, 41(10): 2799-2805. DOI: 10.11772/j.issn.1001-9081.2020121905

    DONG Y F, LIU C, WANG L Q, et al. Relationship reasoning method combining multi-hop relationship path information[J]. Journal of Computer Applications, 2021, 41(10): 2799-2805. DOI: 10.11772/j.issn.1001-9081.2020121905.
    [3] 宋浩楠, 赵刚, 王兴芬. 融合知识表示和深度强化学习的知识推理方法[J]. 计算机工程与应用, 2021, 57(19): 189-197. DOI: 10.3778/j.issn.1002-8331.2104-0430.

    SONG H N, ZHAO G, WANG X F. Knowledge reasoning method combining knowledge representation with deep reinforcement learning[J]. Computer Engineering and Applications, 2021, 57(19): 189-197.DOI: 10.3778/j.issn.1002-8331.2104-0430.
    [4] XU B, XU Y, LIANG J Q, et al. CN-DBpedia: a never-ending Chinese knowledge extraction system[M]//30th International Conference on Industrial Engineering and Other Applications of Applied Intelligent Systems.Arras, France: Springer, 2017: 428-438. DOI: 10.1007%2F978-3-319-60045-1_44.
    [5] BORDES A, USUNIER N, GARCIA-DURN A, et al. Translating embeddings for modeling multi-relational data[C]//Proceedings of the 26th Advances in Neural Information Processing Systems. Red Hook: Curran Associates Inc., 2013: 2787-2795. DOI: 10.5555/2999792.2999923.
    [6] WANG Z, ZHANG J W, FENG J L, et al. Knowledge graph embedding by translating on hyperplanes[C]//Proceedings of the 28th AAAI Conference on Artificial Intelligence. Menlo Park: AAAI, 2014: 1112-1119. DOI: 10.5555/2893873.2894046.
    [7] LIN Y K, LIU Z Y, SUN M S, et al. Learning entity and relation embeddings for knowledge graph completion[C]//Proceedings of the 29th AAAI Conference on Artificial Intelligence. Menlo Park: AAAI, 2015: 2181-2187. DOI: 10.5555/2886521.2886624.
    [8] FENG J, HUANG M L, WANG M D, et al. Knowledge graph embedding by flexible translation[C]//Proceeding of the Fifteenth International Conference on Principles of Knowledge Representation and Reasoning. Cape Town, South Africa: AAAI, 2016: 557-560. DOI: 10.5555/3032027.3032102.
    [9] JI G L, HE S Z, XU L H, et al. Knowledge graph embedding via dynamic mapping matrix[C]//Proceedings of the 53rd Annual Meeting of the Association for Computational Linguistics and the 7th International Joint Conference on Natural Language Processing (Volume 1: Long Papers). Beijing, China: Association for Computational Linguistics, 2015: 687-696. DOI: 10.3115/v1/P15-1067.
    [10] JI G L, LIU K, HE S Z, et al. Knowledge graph completion with adaptive sparse transfer matrix[C]//Proceedings of the Thirtieth AAAI Conference on Artificial Intelligence. Phoenix Arizona: AAAI Press, 2016. DOI: 10.5555/3015812.3015959.
    [11] TAY Y, LUU A T, HUI S C, et al. Random semantic tensor ensemble for scalable knowledge graph link prediction[C]//Proceedings of the 10th ACM International Conference on Web Search and Data Mining. New York: ACM Press, 2017: 751-760. DOI: 10.1145/3018661.3018695.
    [12] XIAO H, HUANG M L, ZHU X Y. From one point to a manifold: Knowledge graph embedding for precise link prediction[C]//Proceedings of the 25th International Joint Conference on Artificial Intelligence. New York: AAAI, 2016: 1315-1321. DOI: 10.5555/3060621.3060804.
    [13] 官赛萍, 靳小龙, 贾岩涛, 等. 面向知识图谱的知识推理研究进展[J]. 软件学报, 2018, 29(10): 2966-2994. DOI: 10.13328/j.cnki.jos.005551.

    GUAN S P, JIN X L, JIA Y T, et al. Knowledge reasoning over knowledge graph: A survey[J]. Journal of Software, 2018, 29(10): 2966-2994. DOI: 10.13328/j.cnki.jos.005551.
    [14] 刘峤, 韩明皓, 杨晓慧, 等. 基于表示学习和语义要素感知的关系推理算法[J]. 计算机研究与发展, 2017, 54(8): 1682-1692. DOI: 10.7544/issn1000-1239.2017.20170200.

    LIU Q, HAN M H, YANG X H, et al. Representation learning based relational inference algorithm with semantical aspect awareness[J]. Journal of Computer Research and Development, 2017, 54(8): 1682-1692. DOI: 10.7544/issn1000-1239.2017.20170200.
    [15] 杨旭东. 基于知识图谱构建的推理补全算法[D]. 成都: 电子科技大学, 2020.

    YANG X D. Inference completion algorithms based on knowledge graph construction[D]. Chengdu: University of Electronic Science and Technology of China, 2020.
    [16] 姜天文, 秦兵, 刘挺. 基于表示学习的开放域中文知识推理[J]. 中文信息学报, 2018, 32(3): 34-41. DOI: 10.3969/j.issn.1003-0077.2018.03.005.

    JIANG T W, QIN B, LIU T. Open domain knowledge reasoning for Chinese based on representation learning[J]. Journal of Chinese Information Processing, 2018, 32(3): 34-41. DOI: 10.3969/j.issn.1003-0077.2018.03.005.
  • 加载中
图(3) / 表(9)
计量
  • 文章访问数:  21
  • HTML全文浏览量:  18
  • PDF下载量:  2
  • 被引次数: 0
出版历程
  • 收稿日期:  2021-09-14
  • 修回日期:  2021-10-21

目录

    /

    返回文章
    返回