Character, Word, or Both? Revisiting the Segmentation Granularity for Chinese Pre-trained Language Models
This work addresses the limitation of existing Chinese PLMs that ignore word information, offering a solution for improved semantic representation in Chinese and Japanese NLP applications, though it is incremental in building upon existing BERT architectures.
The paper tackles the problem of segmentation granularity in Chinese pre-trained language models by proposing MigBERT, which incorporates both character and word-level representations, achieving new state-of-the-art performance on various Chinese NLP tasks and showing effectiveness in Japanese as well.
Pretrained language models (PLMs) have shown marvelous improvements across various NLP tasks. Most Chinese PLMs simply treat an input text as a sequence of characters, and completely ignore word information. Although Whole Word Masking can alleviate this, the semantics in words is still not well represented. In this paper, we revisit the segmentation granularity of Chinese PLMs. We propose a mixed-granularity Chinese BERT (MigBERT) by considering both characters and words. To achieve this, we design objective functions for learning both character and word-level representations. We conduct extensive experiments on various Chinese NLP tasks to evaluate existing PLMs as well as the proposed MigBERT. Experimental results show that MigBERT achieves new SOTA performance on all these tasks. Further analysis demonstrates that words are semantically richer than characters. More interestingly, we show that MigBERT also works with Japanese. Our code and model have been released here~\footnote{https://github.com/xnliang98/MigBERT}.