基于关系触发词与单层GRU模型的关系抽取方法

吉林大学学报(理学版) ›› 2020, Vol. 58 ›› Issue (1): 95-103.

基于关系触发词与单层GRU模型的关系抽取方法

王磊^1, 刘露^{1,2,3,4}, 牛亮^5, 胡封晔^4, 彭涛^1,2

1. 吉林大学计算机科学与技术学院, 长春 130012； 2. 吉林大学符号计算与知识工程教育部重点实验室, 长春 130012； 3. 吉林大学软件学院, 长春 130012； 4. 吉林大学通信工程学院, 长春 130012； 5. 吉林大学第一医院, 长春 130021

收稿日期:2019-06-24 出版日期:2020-01-26 发布日期:2020-01-12
通讯作者: 彭涛 E-mail:tpeng@jlu.edu.cn

Relation Extraction Method Based on Relation Trigger Words and SingleLayer GRU Model

WANG Lei^1, LIU Lu^{1,2,3,4}, NIU Liang^5, HU Fengye^4, PENG Tao^1,2

1. College of Computer Science and Technology, Jilin University, Changchun 130012, China; 2. Key Laboratory of Symbol Computation and Knowledge Engineering for Ministry of Education, Jilin University, Changchun 130012, China; 3. College of Software, Jilin University, Changchun 130012, China;4. College of Communication Engineering, Jilin University, Changchun 130012, China;5. The First Hospital of Jilin University, Changchun 130021, China

Received:2019-06-24 Online:2020-01-26 Published:2020-01-12
Contact: PENG Tao E-mail:tpeng@jlu.edu.cn

摘要/Abstract

摘要： 基于关系触发词与单层门控循环单元模型进行关系抽取, 以降低关系抽取模型结构的复杂度, 并提高模型的训练效率. 通过计算单词的依存距离与序列距离得到关系触发词, 利用单层门控循环单元模型进行关系抽取, 并在SemEval 2010 Task 8数据集上进行实验. 实验结果表明, 该方法能有效提取出关系触发词, 并具有较高的关系抽取准确率.

关键词: 关系抽取, 关系触发词, 句法依存分析, Word2Vec模型, 门控循环单元

Abstract: Relation trigger words and the singlelayer gated recurrent unit model were used for relation extraction in order to reduce the complexity of the relation extraction model structure and improve the training efficiency of the model. By calcul
ating the dependency distance and the sequence distance of the words, we obtained relation trigger words, and used the singlelayer gated recurrent unit model to extract relations. The experiment was performed on the SemEval 2010 Task 8 dataset. The experimental results show that the method can effectively extract the relation trigger words, and has higher accuracy of relation extraction.

Key words: relation extraction, relation trigger word, syntactic dependency parsing, Word2Vec model, gated recurrent unit (GRU)

中图分类号:

TP391

王磊, 刘露, 牛亮, 胡封晔, 彭涛. 基于关系触发词与单层GRU模型的关系抽取方法[J]. 吉林大学学报(理学版), 2020, 58(1): 95-103.

WANG Lei, LIU Lu, NIU Liang, HU Fengye, PENG Tao. Relation Extraction Method Based on Relation Trigger Words and SingleLayer GRU Model[J]. Journal of Jilin University Science Edition, 2020, 58(1): 95-103.

[1]	郑余祥, 左祥麟, 左万利, 梁世宁, 王英. 基于时间关系的Bi-LSTM+GCN因果关系抽取[J]. 吉林大学学报(理学版), 2021, 59(3): 643-648.
[2]	尹伟石, 杨文红, 曲福恒. 无相位远场数据反演散射障碍的神经网络方法[J]. 吉林大学学报(理学版), 2020, 58(6): 1357-1365.
[3]	朱兆龙, 欧阳丹彤, 叶育鑫. Distant Supervision方法中对齐数据的聚类去噪[J]. 吉林大学学报(理学版), 2014, 52(02): 280-284.