arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2106.09233cs.CL
收藏

基于因果干预的远程监督命名实体识别去偏

De-biasing Distantly Supervised Named Entity Recognition via Causal Intervention

University of Chinese Academy of Sciences(中国科学院大学);Institute of Software, Chinese Academy of Sciences(中国科学院软件研究所)

Wenkai Zhang, Hongyu Lin, Xianpei Han, Le Sun

摘要:

远程监督通过字典匹配自动生成训练实例来解决NER中的数据瓶颈。不幸的是,DS-NER的学习是严重的字典偏见,这遭受虚假的相关性,因此破坏了学习模型的有效性和鲁棒性。本文从结构因果模型(SCM)的角度对词典偏倚进行了解释,将其分为词典内偏倚和词典间偏倚,并分析了其产生的原因。基于SCM,我们通过因果干预学习去偏DS-NER。对于字典内偏差,我们进行后门调整,以消除虚假的相关性引入的字典混淆。对于字典间偏差,我们提出了一个因果不变正则化器,这将使DS-NER模型对字典的扰动更具鲁棒性。在四个数据集和三个DS-NER模型上的实验表明,该方法可以显著提高DS-NER的性能。

英文摘要:

Distant supervision tackles the data bottleneck in NER by automatically generating training instances via dictionary matching. Unfortunately, the learning of DS-NER is severely dictionary-biased, which suffers from spurious correlations and therefore undermines the effectiveness and the robustness of the learned models. In this paper, we fundamentally explain the dictionary bias via a Structural Causal Model (SCM), categorize the bias into intra-dictionary and inter-dictionary biases, and identify their causes. Based on the SCM, we learn de-biased DS-NER via causal interventions. For intra-dictionary bias, we conduct backdoor adjustment to remove the spurious correlations introduced by the dictionary confounder. For inter-dictionary bias, we propose a causal invariance regularizer which will make DS-NER models more robust to the perturbation of dictionaries. Experiments on four datasets and three DS-NER models show that our method can significantly improve the performance of DS-NER.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑