一种句袋注意力远程监督关系抽取方法  

A NOVEL DISTANT SUPERVISION RELATION EXTRACTION APPROACH BASED ON SENTENCE BAG ATTENTION

作　　者：张水晶 陈建峡[1] 吴歆韵 Zhang Shuijing;Chen Jianxia;Wu Xinyun(School of Computer Science,Hubei University of Technology,Wuhan 430068,Hubei,China)

机构地区：[1]湖北工业大学计算机学院,湖北武汉430068

出　　处：《计算机应用与软件》2022年第8期193-203,共11页Computer Applications and Software

基　　金：国家自然科学基金青年科学基金项目(61902116)。

摘　　要：为了促进知识图谱技术在课程教学中的应用,解决课程知识点的实体关系抽取困难问题,提出基于句袋注意力的远程监督关系抽取方法。利用远程监督方法从“大数据处理技术”课程教学文本中自动获取训练语料;用PCNN提取句子特征,采用句袋注意力机制对远程监督方法标注的数据中存在的大量噪声去噪;通过带有注意力的词向量捕捉上下文语义信息,并融合实体的位置信息、类型信息构造实体特征,输入Bi_LSTM模型获得知识点关系抽取。实验表明,该方法在“大数据处理技术”课程实现F1值为88.1%的知识点关系抽取。To promote the application of knowledge graph technology in the course teaching and solve the difficult problem that entity relationship extraction of the course knowledge,this paper proposes a novel approach of distant supervision relationship extraction based on sentence-bag attention,named DSRE-SBA.The training corpus was automatically obtained from the teaching texts of the course named"Big Data Processing Technology".Afterward,this approach utilized Piecewise CNN(PCNN)to extract sentence features via a sentence bag attention mechanism to solve the problem of the noisy training data generated by distant supervision.This approach captured the contextual meaning via the word vector with attention.Entity features,combined with the location information and the type information of the entity,were input to the Bi_LSTM model to extract the knowledge relationship.Experiments show that the proposed approach can achieve the F1 value of 88.1%in relationship extraction in the course of"Big Data Processing Technology".

关 键 词：知识图谱 关系抽取 远程监督 句袋模型 注意力机制 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]