Abstract:<p>The success of neural network based methods in named entity recognition (NER) is heavily relied on abundant manual labeled data. However, these NER methods are unavailable when the data is fully-unlabeled in a new domain. To address the problem, we propose an unsupervised cross-domain model which leverages labeled data from source domain to predict entities in unlabeled target domain. To relieve the distribution divergence when transferring knowledge from source to target domain, we apply adversarial training. Furthermore, we design an entity-aware attention module to guide the adversarial training to reduce the discrepancy of entity features between different domains. Experimental results demonstrate that our model outperforms other methods and achieves state-of-the-art performance.</p>

What problem does this paper attempt to address?

The problem that this paper attempts to solve is how to effectively use the labeled data in the source domain to predict entities in the target domain when performing named entity recognition (NER) in a new domain and the data in the target domain is completely unlabeled. Specifically, the application of existing NER methods in new domains depends on a large amount of manually - labeled data, but in many cases, such data are unavailable, especially in some emerging technological fields. Therefore, the paper proposes an unsupervised cross - domain model. This model uses adversarial training and entity - aware attention mechanisms, aiming to reduce the distribution differences between different domains and achieve fine - grained entity feature alignment, thereby improving the entity recognition performance in the target domain. The main contributions of the paper include: 1. Proposing an entity - aware adversarial training model for unsupervised cross - domain NER tasks, which solves the limitations of existing work, including the dependence on external resources and the domain distribution shift problem. 2. Using adversarial training to reduce domain distribution shift and introducing an entity - aware attention mechanism to alleviate the problem of misaligned entity features during the adversarial training process. 3. Evaluating the proposed method through experiments on three datasets. The experimental results show that this model is superior to other existing methods. By combining adversarial training and entity - aware attention mechanisms, the paper effectively solves the key challenges in cross - domain NER tasks, that is, achieving knowledge transfer from the source domain to the target domain without additional large - scale language resources or domain dictionaries.

Unsupervised cross-domain named entity recognition using entity-aware adversarial training

Attention-based Cross-Layer Domain Alignment for Unsupervised Domain Adaptation

Cross-Lingual Named Entity Recognition Based on Attention and Adversarial Training

Neural Adaptation Layers for Cross-domain Named Entity Recognition

Adversarial Transfer Learning for Chinese Named Entity Recognition with Self-Attention Mechanism.

Cross-domain Named Entity Recognition via Graph Matching

Named entity recognition for Chinese based on global pointer and adversarial training

Cross-Domain and Semisupervised Named Entity Recognition in Chinese Social Media: A Unified Model.

A Unified Model for Cross-Domain and Semi-Supervised Named Entity Recognition in Chinese Social Media

Adversarial Transfer Learning for Biomedical Named Entity Recognition

Unsupervised Cross-Lingual Speech Emotion Recognition Using DomainAdversarial Neural Network

Semi-Supervised Disentangled Framework for Transferable Named Entity Recognition

A Double Adversarial Network Model for Multi-Domain and Multi-Task Chinese Named Entity Recognition

Research on Named Entity Recognition for Spoken Language Understanding Using Adversarial Transfer Learning

Enhanced Chinese Domain Named Entity Recognition: An Approach with Lexicon Boundary and Frequency Weight Features

Data Augmentation for Cross-Domain Named Entity Recognition

Named entity recognition based on semi-supervised ensemble learning with the improved tri-training algorithm

Unsupervised Domain Adaptation with Target Reconstruction and Label Confusion in the Common Subspace

Unsupervised Cross-Modality Domain Adaptation of ConvNets for Biomedical Image Segmentations with Adversarial Loss

Using Domain Knowledge for Low Resource Named Entity Recognition

DART: Domain-Adversarial Residual-Transfer Networks for Unsupervised Cross-Domain Image Classification