其他分享
首页 > 其他分享> > 笔记:Simultaneously Self-Attending to All Mentions for Full-Abstract Biological Relation Extraction

笔记:Simultaneously Self-Attending to All Mentions for Full-Abstract Biological Relation Extraction

作者:互联网

Simultaneously Self-Attending to All Mentions for Full-Abstract Biological Relation Extraction

作者:Verga et al., NAACL 2018.

目录

1 简介

主要针对文档级别实体关系抽取,基于Transformer编码。由于针对整个文档中所有的mentions,那么需要跨语句处理,利用muti-instance learning聚合所有的mentions表示,得到span表示,预测关系类型,同时muti-instance learning对于噪声也有一定的抑制作用。

2 模型

关系抽取整体结构如图1所示。

3 总结

看这篇主要是为了补一下(Eberts et al., EACL 2021)\(^{[2]}\)这篇主要参考的paper以及简单了解一下muti-instance learning在doc-level RE的应用。

参考

[1] Patrick Verga, Emma Strubell, Andrew McCallum.Simultaneously Self-Attending to All Mentions for Full-Abstract Biological Relation Extraction.NAACL 2018.

[2] Markus Eberts.Adrian Ulges.An End-to-end Model for Entity-level Relation Extraction using Multi-instance Learning. EACL 2021.

标签:instance,Full,Simultaneously,每个,mention,实体,times,Extraction,token
来源: https://www.cnblogs.com/n-ooo/p/16216227.html