DualNER: A Dual-Teaching framework for Zero-shot Cross-lingual Named Entity Recognition
This work addresses the problem of named entity recognition across languages without target language annotations, which is incremental in combining existing paradigms.
The paper tackles zero-shot cross-lingual named entity recognition by proposing DualNER, a framework that leverages both annotated source language data and unlabeled target language text, achieving improved performance as demonstrated in experiments.
We present DualNER, a simple and effective framework to make full use of both annotated source language corpus and unlabeled target language text for zero-shot cross-lingual named entity recognition (NER). In particular, we combine two complementary learning paradigms of NER, i.e., sequence labeling and span prediction, into a unified multi-task framework. After obtaining a sufficient NER model trained on the source data, we further train it on the target data in a {\it dual-teaching} manner, in which the pseudo-labels for one task are constructed from the prediction of the other task. Moreover, based on the span prediction, an entity-aware regularization is proposed to enhance the intrinsic cross-lingual alignment between the same entities in different languages. Experiments and analysis demonstrate the effectiveness of our DualNER. Code is available at https://github.com/lemon0830/dualNER.