Named entity recognition (NER) suffers from the scarcity of annotated training data, especially for low-resource languages without labeled data. Cross-lingual NER has been proposed to alleviate this issue by transferring knowledge from high-resource languages to low-resource languages via aligned cross-lingual representations or machine translation results. However, the performance of cross-lingual NER methods is severely affected by the unsatisfactory quality of translation or label projection. To address these problems, we propose a Cross-lingual Entity Projection framework (CROP) to enable zero-shot cross-lingual NER with the help of a multilingual labeled sequence translation model. Specifically, the target sequence is first translated into the source language and then tagged by a source NER model. We further adopt a labeled sequence translation model to project the tagged sequence back to the target language and label the target raw sentence. Ultimately, the whole pipeline is integrated into an end-to-end model by the way of self-training. Experimental results on two benchmarks demonstrate that our method substantially outperforms the previous strong baseline by a large margin of +3~7 F1 scores and achieves state-of-the-art performance.

As deep learning architectures are a technique to write a learning system where gradients are the only necessary requirements. FNet uses the Fourier transform to replace the Self-Attention of BERT [3]. The Fourier transform is a technique to embedding an existing function by one using the sinusoidal functions as a basis which originally was though to take O(n²) time complexity where n exists as the size of the input. The Cooley-Tukey Paper from Scripps described a method which takes O(n log n) in 1965 [1]. The Fast Fourier Transform was found because of performing the calculations by hand, a possible reason why people use pen and paper.