PG-HIVE: Hybrid Incremental Schema Discovery for Property Graphs

To test PG-HIVE, you do not need to manually prepare datasets or experiments.

All required datasets, along with detailed instructions for reproducing the experiments, are available here:
https://zenodo.org/records/17801336

The full paper is available here:
https://openproceedings.org/2026/conf/edbt/paper-201.pdf

---

Run PG-HIVE in a few steps

1. Download the datasets from Zenodo and follow the provided instructions.
2. Load a dataset into Neo4j (version 4.4.0).
3. Run the following Cypher query:

   CALL { MATCH (n) SET n.original_label = labels(n) } IN TRANSACTIONS OF 1000 ROWS

4. Navigate to the project:

   cd schemadiscovery

5. Run PG-HIVE:

   sbt "run LSH"

   or for incremental schema discovery:

   sbt "run LSH INCREMENTAL"

---

Citation

If you use PG-HIVE, please cite:

@inproceedings{sideri2026pghive,
  title={PG-HIVE: Hybrid Incremental Schema Discovery for Property Graphs},
  author={Sideri, Sofia and Troullinou, Georgia and Ymeralli, Elisjana and Efthymiou, Vasilis and Plexousakis, Dimitris and Kondylakis, Haridimos},
  booktitle={Proceedings of the EDBT 2026 Conference},
  year={2026},
  url={https://openproceedings.org/2026/conf/edbt/paper-201.pdf}
}