基于知识图谱嵌入的定义域值域约束补全方法

doi:10.13229/j.cnki.jdxbgxb20200755

摘要/Abstract

摘要：

针对知识图谱中定义域值域约束补全问题，将其转化成链接预测问题并使用基于知识图谱嵌入模型预测缺失数据。针对定义域值域约束补全问题的结构，在两种基于翻译的知识图谱嵌入模型TransE和RotatE基础上给出了两种高效的约束补全方法DCaT-T和DCaT-R。特别地，为提升补全方法的预测性能，DCaT-T与DCaT-R方法均采用了两阶段的训练方法。实验结果表明，DCaT-T和DCaT-R优于类型预测方法SDType，DCaT-T方法优于同在TransE基础上实现的嵌入表示模型ConnectE类型预测方法，并且两阶段的训练方法能够进一步提高模型的预测能力。

关键词: 人工智能, 知识图谱, 定义域值域, 约束补全, 知识图谱嵌入

Abstract:

In this paper， we focus on completing the missing domain and range constraints in knowledge graphs and try to predict missing constraints by knowledge graph embedding models. Considering the structure of constraints completion problem， we introduce two efficient approaches， DCaT-T and DCaT-R， which are derived from translation-based knowledge graph embedding models TransE and RotatE. In particular both DCaT-T and DCaT-R exploit a two-stage training approach to improve the performance of the constraints predicting models. Experimental results show that both DCaT-T and DCaT-R are efficient than entity typing approach SDType， DCaT-T performs better than TransE-based entity typing model ConnectE and the two-stage learning approach can improve the performance of the models further.

Key words: artificial intelligence, knowledge graph, domain and range, constraints completion, knowledge graph embedding

中图分类号:

TP391

雷景佩,欧阳丹彤,张立明. 基于知识图谱嵌入的定义域值域约束补全方法[J]. 吉林大学学报(工学版), 2022, 52(1): 154-161.

Jing-pei LEI,Dan-tong OUYANG,Li-ming ZHANG. Relation domain and range completion method based on knowledge graph embedding[J]. Journal of Jilin University(Engineering and Technology Edition), 2022, 52(1): 154-161.

图/表 6

表1

表2

表3

图1

表4

图2

参考文献 18

1	Wang Q, Mao Z, Wang B, et al. Knowledge graph embedding: a survey of approaches and applications[J]. IEEE Transactions on Knowledge and Data Engineering, 2017, 29(12):2724⁃2743.
2	刘知远,孙茂松,林衍凯,等.知识表示学习研究进展[J]. 计算机研究与发展,2016,53(2):247-261.
	Liu Zhi-yuan, Sun Mao-song, Lin Yan-kai, et al. Knowledge representation learning: a review[J]. Journal of Computer Research and Development, 2016, 53(2): 247-261.
3	Krompaß D, Baier S, Tresp V. Type-constrained representation learning in knowledge graphs[C]∥Proceedings of the 14th International Semantic Web Conference, Osaka, Japan,2015: 640-655.
4	Lv X, Hou L, Li J, et al. Differentiating concepts and instances for knowledge graph embedding[C]∥Proceedings of the EMNLP, Brussels, Belgium,2018:1971-1979.
5	Hao J, Chen M, Yu W, et al. Universal representation learning of knowledge bases by jointly embedding instances and ontological concepts[C]∥Proceedings of the 25th ACM SIGKDD, New York,USA,2019:1709-1719.
6	Lei J, Ouyang D, Liu Y. Adversarial knowledge representation learning without external model[J]. IEEE Access, 2017, 29(12):2724⁃2743.
7	Bollacker K, Evans C, Paritosh P, et al. Freebase: a collaboratively created graph database for structuring human knowledge[C]∥ Proceedings of the SIGMOD Conference, San Francisco, USA,2008: 1247-1249.
8	Miller G A. WordNet: a lexical database for English[J]. Communications of the Acm, 1995, 38(11):39-41.
9	Bordes A, Usunier N. Translating embeddings for modeling multi-relational data[C]∥Proceedings of the 27th Annual Conference on Neural Information Processing Systems, Cambridge, USA, 2014:434-449.
10	Sun Z, Deng Z, Nie L, et al. RotatE: knowledge graph embedding by relational rotation in complex space[C]∥ Proceedings of the 7th International Conference on Learning Representations, Bloomington Indiana, USA, 2019: No.978-0-8058-6174-7.
11	Paulheim H, Bizer C. Type inference on noisy RDF data[C]∥Proceedings of the 12th International Semantic Web Conference, Sydney Australia, 2013:510-525.
12	Zhao Y, Zhang A, Xie R, et al. Connecting embeddings for knowledge graph entity typing[C]∥Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, Online,2020:6419-6428.
13	Brickley D, Guha R V. RDF Schema1.1 [DB/OL]. [2020-09-08]. .
14	Lehmann J, Isele R, Jakob M, et al. DBpedia - a large‒scale, multilingual knowledge base extracted from Wikipedia[J]. Semantic Web, 2015, 6(2): 167-195.
15	Liu H, Wu Y, Yang Y. Analogical inference for multi-relational embeddings[C]∥Proceedings of the 34th International Conference on Machine Learning, Sydney,Australia, 2017: 2168-2178.
16	Demers T, Minervini P, Stenetorp P, et al. Convolutional 2D knowledge graph embeddings[C]∥ Proceedings of the 32nd AAAI Conference on Artificial Intelligence, New York, USA, 2018:1811-1818.
17	Zhang S, Tay Y, Yao L, et al. Quaternion knowledge graph embeddings[C]∥ Proceedings of the 33rd Annual Conference on Neural Information Processing Systems, Vancouver,Canada,2019:2731-2741.
18	Liu H, Wu Y, Yang Y. Learn to grow: a continual structure learning framework for overcoming catastrophic forgetting[C]∥ Proceedings of the 36th International Conference on Machine Learning, California, USA, 2019:3925-3934.

相关文章 15

[1]	李志华,张烨超,詹国华. 三维水声海底地形地貌实时拼接与可视化[J]. 吉林大学学报(工学版), 2022, 52(1): 180-186.
[2]	欧阳丹彤,张必歌,田乃予,张立明. 结合格局检测与局部搜索的故障数据缩减方法[J]. 吉林大学学报(工学版), 2021, 51(6): 2144-2153.
[3]	徐艳蕾,何润,翟钰婷,赵宾,李陈孝. 基于轻量卷积网络的田间自然环境杂草识别方法[J]. 吉林大学学报(工学版), 2021, 51(6): 2304-2312.
[4]	杨勇,陈强,曲福恒,刘俊杰,张磊. 基于模拟划分的SP⁃k⁃means-+算法[J]. 吉林大学学报(工学版), 2021, 51(5): 1808-1816.
[5]	赵亚慧,杨飞扬,张振国,崔荣一. 基于强化学习和注意力机制的朝鲜语文本结构发现[J]. 吉林大学学报(工学版), 2021, 51(4): 1387-1395.
[6]	董延华,刘靓葳,赵靖华,李亮,解方喜. 基于BPNN在线学习预测模型的扭矩实时跟踪控制[J]. 吉林大学学报(工学版), 2021, 51(4): 1405-1413.
[7]	朱小龙,谢忠. 基于海量文本数据的知识图谱自动构建算法[J]. 吉林大学学报(工学版), 2021, 51(4): 1358-1363.
[8]	刘富,梁艺馨,侯涛,宋阳,康冰,刘云. 模糊c-harmonic均值算法在不平衡数据上改进[J]. 吉林大学学报(工学版), 2021, 51(4): 1447-1453.
[9]	尚福华,曹茂俊,王才志. 基于人工智能技术的局部离群数据挖掘方法[J]. 吉林大学学报(工学版), 2021, 51(2): 692-696.
[10]	赵海英,周伟,侯小刚,张小利. 基于多任务学习的传统服饰图像双层标注[J]. 吉林大学学报(工学版), 2021, 51(1): 293-302.
[11]	段阳,侯力,冷松. 金属切削加工知识图谱构建及应用[J]. 吉林大学学报(工学版), 2021, 51(1): 122-133.
[12]	欧阳丹彤,马骢,雷景佩,冯莎莎. 知识图谱嵌入中的自适应筛选[J]. 吉林大学学报(工学版), 2020, 50(2): 685-691.
[13]	李贻斌,郭佳旻,张勤. 人体步态识别方法与技术[J]. 吉林大学学报(工学版), 2020, 50(1): 1-18.
[14]	徐谦,李颖,王刚. 基于深度学习的行人和车辆检测[J]. 吉林大学学报(工学版), 2019, 49(5): 1661-1667.
[15]	高万夫,张平,胡亮. 基于已选特征动态变化的非线性特征选择方法[J]. 吉林大学学报(工学版), 2019, 49(4): 1293-1300.

Metrics

Viewed

Full text

Abstract

Cited

Shared

Discussed

数据集	#Entities	#Rel	#Train	#Valid	#Test
FB15k	14 951	1345	483 142	50 000	59 071
WN18	40 943	18	141 442	5000	5000
FB15k?237	14 541	237	272 115	17 535	20 466
WN18RR	40 559	11	86 835	3034	3134

数据集	#C_g（#Dom）	#Train _g	#Valid_g	#Test_g
FB15k	2690	273 710	16 052	19 061
WN18	36	171 254	9006	9028
FB15k?237	474	149 689	5707	6736
WN18RR	22	103 509	2741	2811

Data set	metric	SDType	DCaT-T	DCaT-R
FB15k	MR	86.42	3.77	3.95
	MRR	0.460	0.825	0.815
	Hits@1	0.366	0.761	0.746
	Hits@3	0.496	0.869	0.863
	Hits@10	0.696	0.946	0.944
WN18	MR	1.61	1.15	1.19
	MRR	0.936	0.976	0.968
	Hits@1	0.912	0.965	0.955
	Hits@3	0.968	0.983	0.977
	Hits@10	0.993	0.996	0.996
FB15k?237	MR	10.7	5.340	5.52
	MRR	0.502	0.543	0.545
	Hits@1	0.365	0.365	0.375
	Hits@3	0.567	0.659	0.649
	Hits@10	0.792	0.892	0.879
WN18RR	MR	2.60	2.00	1.70
	MRR	0.771	0.802	0.828
	Hits@1	0.702	0.716	0.729
	Hits@3	0.784	0.860	0.915
	Hits@10	0.901	0.990	0.993

metric	ConnectE	DCaT?T	DCaT?T （N）	DCaT?T （MIX）
MR	14.65	3.770	3.79	4.18
MRR	0.52	0.825	0.822	0.806
Hits@1	0.397	0.761	0.755	0.734
Hits@3	0.573	0.869	0.868	0.859
Hits@10	0.787	0.946	0.945	0.943