摘要
针对现有中文因果关系抽取方法对因果事件边界难以识别和文本特征表示不充分的问题,该文提出了一种基于外部词汇信息和注意力机制的中文因果关系抽取模型BiLSTM-TWAM+CRF。该模型使用SoftLexicon方法引入外部词汇信息构建词集,解决了因果事件边界难以识别的问题。通过构建的双路关注模块TWAM(Two Way Attention Module),实现了从局部和全局两个角度充分刻画文本特征。实验结果表明,与当前中文因果关系抽取模型相比较,该文所提方法表现出更优的抽取效果。
Existing Chinese causality extraction methods is challenged by the issue of causal event boundaries identifi-cation and inadequate text features representation.This paper proposes a Chinese causality extraction model BiL-STM-TWAM+CRF which ustilizes external lexical information and attention mechanism.We introduce the external lexical information by using the SoftLexicon to construct word set for determining causal event boundaries.We con-structe a Two Way Attention Module(TWAM)to represent text features from both the local and global views.Ex-perimental results show that our proposed method has better performance than the existing Chinese causality extrac-tion methods.
作者
崔仕林
闫蓉
CUI Shilin;YAN Rong(College of Computer Science,Inner Mongolia University,Hohhot,Inner Mongolia 010021,China;Inner Mongolia Key Laboratory of Mongolian Information Processing Technology,Hohhot,Inner Mongolia 010021,China;National and Local Joint Engineering Research Center of Intelligent Information Processing Technology for Mongolian,Hohhot,Inner Mongolia 010021,China)
出处
《中文信息学报》
CSCD
北大核心
2023年第4期81-89,共9页
Journal of Chinese Information Processing
基金
国家自然科学基金(61866029)。
关键词
因果关系抽取
序列标注
注意力机制
causal relation extraction
sequence labeling
attention mechanism
