Revisiting Pre-Trained Models for Chinese Natural Language Processing

Abstract

Bidirectional Encoder Representations from Transformers (BERT) has shown marvelous improvements across various NLP tasks, and consecutive variants have been proposed to further improve the performance of the pre-trained language models. In this paper, we target on revisiting Chinese pre-trained language models to examine their effectiveness in a non-English language and release the Chinese pre-trained language model series to the community. We also propose a simple but effective model called MacBERT, which improves upon RoBERTa in several ways, especially the masking strategy that adopts MLM as correction (Mac). We carried out extensive experiments on eight Chinese NLP tasks to revisit the existing pre-trained language models as well as the proposed MacBERT. Experimental results show that MacBERT could achieve state-of-the-art performances on many NLP tasks, and we also ablate details with several findings that may help future research. Resources available: https://github.com/ymcui/MacBERT

arXiv comment12 pages, to appear at Findings of EMNLP 2020

Cited by in corpus (13)

• Towards artificial general intelligence via a multimodal foundation model
• CPT: A Pre-Trained Unbalanced Transformer for Both Chinese Language Understanding and Generation
• Survey of different Large Language Model Architectures: Trends, Benchmarks, and Challenges
• Bringing order into the realm of Transformer-based language models for artificial intelligence and law
• When Geoscience Meets Foundation Models: Towards General Geoscience Artificial Intelligence System
• Non-autoregressive Transformer-based End-to-end ASR using BERT
• More but Correct: Generating Diversified and Entity-revised Medical Response
• Machine Learning for Violence Risk Assessment Using Dutch Clinical Notes
• FuDFEND: Fuzzy-domain for Multi-domain Fake News Detection
• FaultProfIT: Hierarchical Fault Profiling of Incident Tickets in Large-scale Cloud Systems
• SZTU-CMU at MER2024: Improving Emotion-LLaMA with Conv-Attention for Multimodal Emotion Recognition
• A Survey on Awesome Korean NLP Datasets
• AF Adapter: Continual Pretraining for Building Chinese Biomedical Language Model