A scoreboard for Chinese NLP progress

What it does

This repository acts as a centralized index for Chinese NLP benchmarks. It links out to per-task documents covering everything from word segmentation and sentiment analysis to entity linking and machine translation, collecting shared tasks, datasets, and reported state-of-the-art results in one place. Think of it as a field guide to who’s ahead on which Chinese-language benchmark and what data they used to get there.

The interesting bit

The project is essentially curation, not code — a human-maintained wiki of leaderboard links and paper results rather than a framework or toolkit. That makes it useful as a quick reference, though the actual tables live in the linked docs/ files rather than the README itself.

Key highlights

• Covers twenty task areas including coreference resolution, constituency parsing, dialogue state management, and simplified/traditional Chinese conversion
• Maintained by the DiDi NLP team under DiDi AI Labs
• Explicitly positions itself as a Chinese-language companion to the broader nlpprogress.com tracker
• Includes historical shared-task announcements, such as a WMT 2021 triangular machine-translation challenge using English to improve Russian-to-Chinese systems

Caveats

• The README is mostly a table of contents; the actual benchmark data and SOTA tables reside in separate markdown files not rendered in the main landing page
• The “What’s new” section still lists WMT 2021 shared-task deadlines that are now several years out of date

Verdict

Worth bookmarking if you’re doing literature review or benchmark shopping in Chinese NLP. Skip it if you’re looking for training code, model weights, or an actively updated daily leaderboard.