计算机与数字工程2023，Vol.51Issue(11)：2568-2572,2601,6.DOI:10.3969/j.issn.1672-9722.2023.11.018
融合GRU和注意力机制的图卷积关系抽取
Graph Convolution Relation Extraction Based on GRU and Attention Mechanism
杜琰 1孙弋1
作者信息
- 1. 西安科技大学通信与信息工程学院 西安 710054
- 折叠
摘要
Abstract
Entity relation extraction is very important in natural language processing.Aiming at the problems of inaccurate fea-ture extraction in graph convolution network and fuzzy gradient of cyclic neural network,a graph convolution relation extraction mod-el integrating gated cyclic unit(GRU)and attention mechanism is proposed.By adding two-way GRU to process the input context information,more detailed features can be obtained,so as to learn the long-term dependent information,and the multi head atten-tion mechanism is further used to distribute the weight of different types of edges and nodes,filter the redundant information and en-hance the correlation between nodes.Finally,graph convolution is used to get the final relationship extraction result.Experiments on SemEval-2010Task 8 and SemEval-2010Task 4 data sets show that this method improves its F1 value and can effectively extract relationships.关键词
关系抽取/门控循环单元/注意力机制/图卷积网络Key words
relationship extraction/gating loop unit/attention mechanism/graph convolutional network分类
通用工业技术引用本文复制引用
杜琰,孙弋..融合GRU和注意力机制的图卷积关系抽取[J].计算机与数字工程,2023,51(11):2568-2572,2601,6.
