HALvest-Geometric

Citation Network of Open Scientific Papers Harvested from HAL

---

Dataset Summary

overview:

French and English fulltexts from open papers found on Hyper Articles en Ligne (HAL) and its citation network.

You can download the dataset using Hugging Face datasets:

from datasets import load_dataset

ds = load_dataset("Madjakul/HALvest-Geometric", "en")

Details

Nodes

• Papers: 18,662,037
• Authors: 238,397
• Affiliations: 96,105
• Domains: 16

Edges

• Paper <-> Domain: 136,700
• Paper <-> Paper: 22,363,817
• Author <-> Paper: 238,397
• Author <-> Affiliation: 426,030

Languages

ISO-639	Language	# Documents	# mT5 Tokens
en	English	442,892	7,606,895,258
fr	French	193,437	8,728,722,255

Considerations for Using the Data

The corpus is extracted from the HAL's open archive which distributes scientific publications following open access principles. The corpus is made up of both creative commons licensed and copyrighted documents (distribution authorized on HAL by the publisher). This must be considered prior to using this dataset for any purpose, other than training deep learning models, data mining etc. We do not own any of the text from which these data has been extracted.

Dataset Copyright

The licence terms for HALvest strictly follows the one from HAL. Please refer to the below license when using this dataset.

• HAL license

Citation

@misc{kulumba2024harvestingtextualstructureddata,
      title={Harvesting Textual and Structured Data from the HAL Publication Repository}, 
      author={Francis Kulumba and Wissam Antoun and Guillaume Vimont and Laurent Romary},
      year={2024},
      eprint={2407.20595},
      archivePrefix={arXiv},
      primaryClass={cs.DL},
      url={https://arxiv.org/abs/2407.20595}, 
}