MoonNER

MoonNER is a lightweight MoonBit named entity recognition engine for Chinese and mixed-script text.

Repositories

• GitHub: https://github.com/ljlljl678/MoonNER
• GitLink: https://gitlink.org.cn/ljlljl678/moonner

Project name

MoonNER

Project identifier

ljlljl678/moonner

Project summary

MoonNER turns raw text into entity spans with a small, readable pipeline:

• Unicode-aware tokenization with longest-match Chinese phrase splitting
• Lexicon-driven feature scoring for people, organizations, locations, and time expressions
• Linear-chain decoding for stable entity spans
• A thin CLI entrypoint for quick demos

Supported entities

• Person
• Organization
• Location
• Time

Why this project

NER is a good contest topic because it is a mature NLP baseline with room to grow. The current version focuses on a clear, extensible core rather than a narrow proof of concept. That leaves space for later improvements such as:

• richer transition scoring
• training data support
• gazetteer expansion
• evaluation metrics
• web or service integration

Run

moon test
moon run cmd/main -- "浠婂ぉ寮犱紵鍦ㄥ寳浜ぇ瀛﹀拰娣卞湷鐮旂┒闄㈣璁?MoonBit NER 椤圭洰銆?

Structure

• moonner.mbt: public library facade
• ner_types.mbt: entities, tags, and model data
• tokenizer.mbt: text segmentation
• model.mbt: scoring, decoding, and span merging
• cmd/main/main.mbt: CLI entrypoint

License

Apache-2.0

Source note

This repository is original MoonBit code written for OSC2026. The lexicon entries are small hand-written examples for demonstration and are not copied from any external dataset.