Reproduce “Toward Fast and Accurate Neural Chinese Word Segmentation with Multi-Criteria Learning” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Weipeng Huang, Xingyi Cheng, Kunlong Chen, Taifeng Wang, Wei Chu
The authors introduce a Chinese word segmentation system for learning across datasets with boundary criteria. It uses a pretrained BERT encoder, shared and dataset-specific projections, and a CRF. In a three-layer comparison, multi-criteria training raises average F1 from 94.5% to 96.8% and OOV recall from 74.2% to 83.6%, though OOV recall falls on AS. This supports multi-criteria learning, not projection alone.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.