摘要
金融领域文本往往包含大量因果信息,对金融文本进行信息抽取,可以帮助获取大量含有因果关系的事件对,进而更好地应用在各个下游任务中。通过构建金融领域的因果标注数据集,并在BERT模型的基础上进行改进,提出了pipeline结构的因果事件抽取模型PUBERT,获得了较好的抽取效果,对相关抽取研究具有一定的参考价值。
Text in the financial field often contains a large amount of causal information.Information extraction of financial text can help us obtain a large number of event pairs with causal relationships,and then apply them to various downstream tasks.This work proposes pipeline structured causal event extraction model PUBERT by building a causal annotation dataset in the finan⁃cial field and improving on the BERT model,and obtains a good extraction effect,which has certain reference value for relevant ex⁃traction research.
作者
席建文
Xi Jianwen(School of Electronic Information,Southwest Minzu University,Chengdu 610041)
出处
《现代计算机》
2023年第3期97-101,107,共6页
Modern Computer
基金
西南民族大学中央高校基本科研业务费专项资金项目(2021NYYXS77)。
关键词
金融领域
信息抽取
因果关系
financial field
information extraction
causal relationship
