北京大学学报自然科学版 ›› 2020, Vol. 56 ›› Issue (1): 39-44.DOI: 10.13209/j.0479-8023.2019.101

上一篇    下一篇

融合门控机制的远程监督关系抽取方法

李兴亚, 陈钰枫, 徐金安, 张玉洁   

  1. 北京交通大学计算机与信息技术学院, 北京 100044
  • 收稿日期:2019-05-23 修回日期:2019-09-25 出版日期:2020-01-20 发布日期:2020-01-20
  • 通讯作者: 陈钰枫, E-mail: chenyf(at)bjtu.edu.cn
  • 基金资助:
    国家自然科学基金(61976016, 61473294, 61370130, 61876198)、北京市自然科学基金(4172047)和科学技术部国际科技合作计划(K11F100010)资助

Distant Supervision for Relation Extraction with Gate Mechanism

LI Xingya, CHEN Yufeng, XU Jin’an, ZHANG Yujie   

  1. School of Computer and Information Technology, Beijing Jiaotong University, Beijing 100044
  • Received:2019-05-23 Revised:2019-09-25 Online:2020-01-20 Published:2020-01-20
  • Contact: CHEN Yufeng, E-mail: chenyf(at)bjtu.edu.cn

摘要:

提出一种融合门控机制的远程监督关系抽取方法。首先在词级别上自动选择正相关特征, 过滤与关系标签无关的词级别噪声; 然后在门控机制内引入软标签的思想, 弱化硬标签对噪声过滤的影响; 最后结合句子级别的噪声过滤, 提升模型的整体性能。在公开数据集上的实验结果表明, 相对于句子级别噪声过滤方法, 所提方法的性能有显著提高。

关键词: 关系抽取, 远程监督, 门控机制, 卷积神经网络

Abstract:

A piecewise convolutional neural network with gating mechanism is proposed, which would automatically filter positive correlation features at word-level. Moreover, the idea of soft-label is introduced to the gating mechanism to weaken the impact of hard labels on noise filtering. Combined with sentence-level noise filtering, the overall performance of the model is improved. The experimental results on the public dataset show that the proposed model has a significant improvement compared to the sentence-level noise filtering methods. 

Key words: relation extraction, distant supervision, gate mechanism, convolutional neural network