Reproduce “From Global Alignment to Local Grounding: Zero-Shot Chinese Character Recognition with Radical Verification” 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
Yu-Heng Shih, Bing-Chen Wu, Tsz-To Wong, Ting-En Yen, Hong-Han Shuai, Bin-Hua Hsieh, Chien-An Chen, Yi-Ren Yeh and 1 more
The authors introduce a zero-shot method for recognizing handwritten Chinese character categories without training examples for those categories. It enriches radical-and-layout descriptions with spatial geometry to retrieve candidates, then uses local radical evidence to rerank ambiguous cases. With 2,755 seen categories, it reaches 83.06% top-1 accuracy, versus 78.96% for FT-CLIP; gains do not reliably extend to characters with unseen radicals.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.