What it does

AgriKG is a Neo4j-backed knowledge graph for Chinese agriculture, built as a Shanghai government research project. It scrapes Hudong Baike (China’s Wikipedia clone) and Wikidata for entities, runs KNN classification on ~150k items, and wraps everything in a Django web app for entity search, relation queries, and basic Q&A.

The interesting bit

The project openly admits it is “stopped maintenance” — a refreshingly honest epitaph. What’s left is a surprisingly complete snapshot of 2018-era Chinese NLP tooling: Scrapy crawlers, manual annotation tools, Cypher import scripts, and even a shortest-path query feature to find “strange implicit relationships” between entities. The data is free for academic use.

Key highlights

• ~15,000 human-labeled entities plus 150k KNN-predicted labels
• Pre-built CSVs for Neo4j import (no crawling required to start)
• Relation queries support 6 different triple patterns (entity-relation-entity combinations)
• Includes climate-to-plant and city-to-climate mappings
• Published at DASFAA 2019 with a proper citation

Caveats

• Explicitly unmaintained; code is “for reference only”
• Setup requires Neo4j + Django + manual Cypher imports — not a quick demo
• Some data processing scripts are marked “already useless” in the source

Verdict

Grab the CSVs and Cypher scripts if you’re building a Chinese-domain knowledge graph and need starter data. Skip it if you want a maintained, production-ready framework — this is research archaeology, not a product.