Datasets:

Modalities:
Text
Formats:
parquet
Languages:
French
ArXiv:
Libraries:
Datasets
pandas
License:
imenelydiaker's picture
Update README.md (#2)
f941c63 verified
|
raw
history blame
1.41 kB
metadata
dataset_info:
  - config_name: documents
    features:
      - name: text
        dtype: string
      - name: doc_id
        dtype: string
    splits:
      - name: test
        num_bytes: 9476376
        num_examples: 2556
    download_size: 4995857
    dataset_size: 9476376
  - config_name: queries
    features:
      - name: query
        dtype: string
      - name: positive
        sequence: string
      - name: negative
        sequence: string
    splits:
      - name: test
        num_bytes: 1435051
        num_examples: 2316
      - name: train
        num_bytes: 5773011
        num_examples: 9264
    download_size: 2617958
    dataset_size: 7208062
configs:
  - config_name: documents
    data_files:
      - split: test
        path: documents/test-*
  - config_name: queries
    data_files:
      - split: test
        path: queries/test-*
      - split: train
        path: queries/train-*
license: mit
language:
  - fr
pretty_name: Alloprof

Description

This dataset was built upon Alloprof Q&A dataset, negative samples were created using BM25. Please refer to our paper for more details.

Citation

If you use this dataset in your work, please consider citing:

@misc{ciancone2024extending,
      title={Extending the Massive Text Embedding Benchmark to French}, 
      author={Mathieu Ciancone and Imene Kerboua and Marion Schaeffer and Wissam Siblini},
      year={2024},
      eprint={2405.20468},
      archivePrefix={arXiv},
      primaryClass={cs.CL}
}