软件导刊2025，Vol.24Issue(9)：41-47,7.DOI:10.11907/rjdk.241626
基于共享特征的联合实体关系抽取网络
Shared Feature-Based Joint Entity and Relation Extraction Network
摘要
Abstract
In the information extraction process,joint entity relationship extraction plays a key role,which integrates two core sub tasks:named entity recognition and relationship extraction.The current joint extraction methods still suffer from feature misclassification and insuffi-cient interaction between subtasks.Therefore,a joint entity relationship extraction network SJERE based on shared features is proposed.This network adopts a parallel encoding method to obtain named entity feature representations and relationship feature representations,effectively avoiding feature overlap and feature misclassification problems.To achieve bidirectional information exchange between two subtasks,the SJERE network constructed a shared feature module.This module allows the model to assist in relationship prediction through entity informa-tion,thereby promoting and enhancing the two sub tasks of entity recognition and relationship extraction.Compared with existing joint models on three benchmark datasets,NYT,WebNLG,and SciERC,the SJERE network exhibits certain performance advantages.关键词
信息抽取/联合实体关系抽取/并行编码/共享特征/双向交互Key words
information extraction/joint entity and relation extraction/parallel encoding/shared feature/bidirectional interaction分类
信息技术与安全科学引用本文复制引用
郝增昊,赵晶..基于共享特征的联合实体关系抽取网络[J].软件导刊,2025,24(9):41-47,7.基金项目
山东省自然科学基金—联合基金项目(ZR2022LZH008) （ZR2022LZH008）
齐鲁工业大学(山东省科学院)研究生教改项目(YJG23YB002) （山东省科学院）
山东省本科教学改革研究项目重点项目(Z2023314) （Z2023314）
济南市市校融合发展战略工程项目(JNSX2023045) （JNSX2023045）
