Distantly-Supervised Named Entity Recognition with Noise-Robust Learning and Language Model Augmented Self-Training
Distantly-Supervised Named Entity Recognition with Noise-Robust Learning and Language Model Augmented Self-Training
复制标题
DOI：
10.18653/v1/2021.emnlp-main.810
复制
发表时间：
2021-09
期刊：
影响因子：
--
通讯作者：
Yu Meng;Yunyi Zhang;Jiaxin Huang;Xuan Wang;Yu Zhang;Heng Ji;Jiawei Han
中科院分区：
文献类型：
--
作者：
Yu Meng;Yunyi Zhang;Jiaxin Huang;Xuan Wang;Yu Zhang;Heng Ji;Jiawei Han
我们研究的问题，训练命名实体识别（NER）模型，只使用距离标记的数据，这可以自动获得匹配的实体提到的原始文本中的实体类型的知识库。远程监督NER的最大挑战是，远程监督可能会导致不完整和嘈杂的标签，使监督学习的直接应用无效。在本文中，我们提出了（1）一个噪声鲁棒的学习计划，包括一个新的损失函数和一个嘈杂的标签去除步骤，用于训练远距离标记数据的NER模型，和（2）一个自训练方法，使用由预训练的语言模型创建的上下文增强，以提高NER模型的泛化能力。在三个基准数据集上，我们的方法实现了上级性能，显著优于现有的远程监督NER模型。
We study the problem of training named entity recognition (NER) models using only distantly-labeled data, which can be automatically obtained by matching entity mentions in the raw text with entity types in a knowledge base. The biggest challenge of distantly-supervised NER is that the distant supervision may induce incomplete and noisy labels, rendering the straightforward application of supervised learning ineffective. In this paper, we propose (1) a noise-robust learning scheme comprised of a new loss function and a noisy label removal step, for training NER models on distantly-labeled data, and (2) a self-training method that uses contextualized augmentations created by pre-trained language models to improve the generalization ability of the NER model. On three benchmark datasets, our method achieves superior performance, outperforming existing distantly-supervised NER models by significant margins.