Yangtze River Delta Geographic Entity and Relation Data

Zongjun Wei's dataset supports the LLM-Reflex-GeoKG framework for constructing Geographic Knowledge Graphs. It contains data from the Yangtze River Delta river network used to validate automated entity recognition and relation extraction. The dataset is a 5.5 KB XLS file published on figshare in April 2026.

Use Cases

• Benchmark geographic entity recognition models using the provided F1 score of 0.898 as a target.
• Evaluate relation extraction algorithms against the reported baseline performance of 0.823 F1 score.
• Train or test dual-LLM generator-critic schemes for complex geospatial semantics understanding.
• Apply multi-stage extraction strategies for automated knowledge acquisition in river network modelling.

Strengths

• Framework achieved a high F1 score of 0.898 for geographic entity recognition.
• Reported a significant performance improvement over baselines for relation extraction (0.823 F1).

Limitations

• Dataset is extremely small at 5.5 KB, indicating very limited scope and sample size.
• Specific row count, column names, and data structure are unknown.
• Geographic coverage is restricted to the Yangtze River Delta river network.

Provenance

Source

Zongjun Wei

Collection Method

Data used for experiments validating the LLM-Reflex-GeoKG framework.

Freshness

Last updated in April 2026.

Geography

Yangtze River Delta river network