Adversarial Training Improved Multi-Path Multi-Scale Relation Detector for Knowledge Base Question Answering

被引:3
作者
Zhang, Yanan [1 ,2 ,3 ]
Xu, Guangluan [1 ,3 ]
Fu, Xingyu [1 ,3 ]
Jin, Li [1 ,3 ]
Huang, Tingle [4 ]
机构
[1] Chinese Acad Sci, Inst Elect, Beijing 100190, Peoples R China
[2] Univ Chinese Acad Sci, Sch Elect Elect & Commun Engn, Beijing 100190, Peoples R China
[3] Chinese Acad Sci, Inst Elect, Key Lab Network Informat Syst Technol NIST, Beijing 100190, Peoples R China
[4] Chinese Acad Sci, Inst Software, Beijing 100190, Peoples R China
关键词
Adversarial training; knowledge base question answering; relation detection;
D O I
10.1109/ACCESS.2020.2984393
中图分类号
TP [自动化技术、计算机技术];
学科分类号
0812 ;
摘要
Knowledge Base Question Answering (KBQA) is a promising approach for users to access substantial knowledge and has become a research focus in recent years. Our paper focuses on relation detection, a subtask of KBQA and proposes an adversarial training improved multi-path multi-scale relation detector (AdvT-MMRD) to improve the performance of a common KBQA system. To solve the problem of matching the casual form of a question with the logical form of a predicate, we use question pattern-relation matching, in which an attention-based bidirectional recurrent neural network with gated recurrent units (Bi-GRUs) is used to match semantic similarity and a convolutional neural network (CNN) is used to learn literal similarity between question and relation. We also explore two ways to measure the relevance of entity type-relation pairs through several level representations. Additionally, an adversarial training strategy is conducted to enhance our model. The experimental results demonstrate that our approach not only achieves a state-of-the-art accuracy of 93.8 & x0025; on relation detection task, but contributes our KBQA system to reaching an outstanding accuracy of 79.0 & x0025; on the SimpleQuestions benchmark.
引用
收藏
页码:63310 / 63319
页数:10
相关论文
共 46 条
  • [1] [Anonymous], 2018, ARXIV180502917
  • [2] [Anonymous], ARXIV170408384
  • [3] [Anonymous], 2018, ARXIV180403317
  • [4] [Anonymous], 2017, P ACL
  • [5] [Anonymous], 2016, ARXIV160300957
  • [6] [Anonymous], 2016, ABS160600979 CORR
  • [7] [Anonymous], 2018, ARXIV180804126
  • [8] Bast H., 2015, CIKM, P1431, DOI [10.1145/2806416.2806472, DOI 10.1145/2806416.2806472]
  • [9] Bekoulis G., 2018, ARXIV180806876
  • [10] Berant J, 2013, P 2013 C EMP METH NA, P1533