基于森林的实体关系联合抽取模型  

Forest-based entity-relation joint extraction model

作　　者：王炫力 靳小龙[1,2] 侯中妮 廖华明 张瑾[1,2] WANG Xuanli;JIN Xiaolong;HOU Zhongni;LIAO Huaming;ZHANG Jin(Key Laboratory of Network Data Science and Technology,Chinese Academy of Sciences(Institute of Computing Technology,Chinese Academy of Sciences),Beijing 100190,China;University of Chinese Academy of Sciences,Beijing 100049,China)

机构地区：[1]中国科学院网络数据科学与技术重点实验室(中国科学院计算技术研究所),北京100190 [2]中国科学院大学,北京100049

出　　处：《计算机应用》2023年第9期2700-2706,共7页journal of Computer Applications

摘　　要：嵌套实体对实体关系联合提取任务提出了挑战。现有的联合抽取模型在处理嵌套实体时存在产生大量负例且复杂度高的问题,此外未考虑嵌套实体对三元组预测的干扰。针对以上问题,提出一种基于森林的实体关系联合抽取方法——EF2LTF(Entity Forest to Layering Triple Forest)。EF2LTF采用了一个两阶段的联合训练框架,首先通过生成实体森林灵活地在嵌套实体内部识别不同的实体;然后结合已识别出的嵌套实体及其层次结构生成分层的三元组森林。在四个标准数据集上的实验结果表明,与基于集合预测网络的SPN(Set Prediction Network)模型、基于跨度的实体关系联合抽取模型SpERT(Span-based Entity and Relation Transformer)和动态图增强信息抽取(DyGIE++)等方法相比,所提方法取得了最优的F1值。说明所提方法既增强了嵌套实体的识别能力,也增强了构建三元组时对嵌套实体的分辨能力,从而提升了实体与关系的联合抽取性能。Nested entities pose a challenge to the task of entity-relation joint extraction.The existing joint extraction models have the problems of generating a large number of negative examples and high complexity when dealing with nested entities.In addition,the interference of nested entities on triplet prediction is not considered by these models.To solve these problems,a forest-based entity-relation joint extraction method was proposed,named EF2LTF(Entity Forest to Layering Triple Forest).In EF2LTF,a two-stage joint training framework was adopted.Firstly,through the generation of an entity forest,different entities within specific nested entities were identified flexibly.Then,the identified nested entities and their hierarchical structures were combined to generate a hierarchical triplet forest.Experimental results on four benchmark datasets show that EF2LTF outperforms methods such as joint entity and relation extraction with Set Prediction Network(SPN)model,joint extraction model for entities and relations based on Span—SpERT(Span-based Entity and Relation Transformer)and Dynamic Graph Information Extraction++(DyGIE++)on F1 score.It is verified that the proposed method not only enhances the recognition ability of nested entities,but also enhances the ability to distinguish nested entities when constructing triples,thereby improving the joint extraction performance of entities and relations.

关 键 词：实体关系联合抽取 三元组生成 嵌套实体 分层预测 实体森林 

分 类 号：TP391.1[自动化与计算机技术—计算机应用技术]