基于关系触发词与单层GRU模型的关系抽取方法  

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

作　　者：王磊 刘露[1,2,3,4] 牛亮 胡封晔[4] 彭涛[1,2] WANG Lei;LIU Lu;NIU Liang;HU Fengye;PENG Tao(College of Computer Science and Technology,Jilin University,C hangchun 130012,China;Key Laboratory of Symbol Computation and Knowledge Enginee ring for Ministry of Education,Jilin University,Changchun 130012,China;College of Software,Jilin University,Changchun 130012,China;College of Communication Engineering,Jilin Unive rsity,Changchun 130012,China;The First Hospital of Jilin University,Changchun 130021,China)

机构地区：[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

出　　处：《吉林大学学报（理学版）》2020年第1期95-103,共9页Journal of Jilin University:Science Edition

基　　金：国家自然科学基金（批准号：61872163;61806084）;吉林省教育厅“十三五”科学技术研究项目（批准号：JJKH20190160KJ）

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SemEval2010Task8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.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 mod el to extract relations.The experiment was performed on the SemEval 2010 Task 8 dataset.The experimental results show that the method can effectively ex tract the relation trigger words,and has higher accuracy of relation extraction.

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 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]