Datasets:

CATIE-AQ
/

frenchSIMPLIFICATION

dataset_info:
  features:
    - name: sentence
      dtype: string
    - name: simplification
      dtype: string
    - name: dataset
      dtype: string
  splits:
    - name: train
      num_bytes: 339019973
      num_examples: 781801
    - name: validation
      num_bytes: 972654
      num_examples: 2385
    - name: test
      num_bytes: 753090
      num_examples: 1439
  download_size: 218816945
  dataset_size: 340745717
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*
      - split: validation
        path: data/validation-*
      - split: test
        path: data/test-*
language:
  - fra
task_categories:
  - text-generation
pretty_name: frenchSIMPLIFICATION

Dataset information

Dataset concatenating Simplification datasets, available in French and open-source.
There are a total of 785,625 rows, of which 781,801 are for training, 2,385 for validation and 1,439 for testing.

Usage

from datasets import load_dataset
dataset = load_dataset("CATIE-AQ/frenchSIMPLIFICATION")

Dataset

Details of rows

Dataset Original	Splits	Note
clear	4,196 train / 300 validation / 100 test
wikilarge	296,402 train / 992 validation / 359 test
GEM/BiSECT	491,035 train / 2,400 validation / 1,036 test	We keep only the data in French (`fr`)
alector	1,108 train	We cut the 79 original texts into sentences to obtain 1,108 data instead of 79.

Removing duplicate data and leaks

The sum of the values of the datasets listed here gives the following result:

DatasetDict({
    train: Dataset({
        features: ['sentence', 'simplification', 'dataset'],
        num_rows: 792741
    })
    validation: Dataset({
        features: ['sentence', 'simplification', 'dataset'],
        num_rows: 3692
    })
    test: Dataset({
        features: ['sentence', 'simplification', 'dataset'],
        num_rows: 1495
    })
})

However, a data item in training split A may not be in A's test split, but may be present in B's test set, creating a leak when we create the A+B dataset.
The same logic applies to duplicate data. So we need to make sure we remove them.
After our clean-up, we finally have the following numbers:

DatasetDict({
    train: Dataset({
        features: ['sentence', 'simplification', 'dataset'],
        num_rows: 781801
    })
    validation: Dataset({
        features: ['sentence', 'simplification', 'dataset'],
        num_rows: 2385
    })
    test: Dataset({
        features: ['sentence', 'simplification', 'dataset'],
        num_rows: 1439
    })
})

Columns

the sentence column contains the text
the simplification column contains the simplification of the sentence
the dataset column identifies the row's original dataset (if you wish to apply filters to it)

Split

train corresponds to the concatenation of clear + wikilarge + bisect + alector
validation corresponds to the concatenation of clear + wikilarge + bisect
test corresponds to clear + wikilarge + bisect

Citations

Alector

@inproceedings{gala-etal-2020-alector,
    title = "{A}lector: A Parallel Corpus of Simplified {F}rench Texts with Alignments of Misreadings by Poor and Dyslexic Readers",  
    author = {Gala, N{\'u}ria and Tack, Ana{\"\i}s  and Javourey-Drevet, Ludivine and Fran{\c{c}}ois, Thomas and Ziegler, Johannes C.},  
    editor = "Calzolari, Nicoletta and B{\'e}chet, Fr{\'e}d{\'e}ric and Blache, Philippe  and Choukri, Khalid and Cieri, Christopher and Declerck, Thierry and Goggi, Sara  and Isahara, Hitoshi and Maegaard, Bente and Mariani, Joseph and Mazo, H{\'e}l{\`e}ne and Moreno, Asuncion  and Odijk, Jan  and Piperidis, Stelios",  
    booktitle = "Proceedings of the Twelfth Language Resources and Evaluation Conference",  
    month = may,  
    year = "2020",  
    address = "Marseille, France",  
    publisher = "European Language Resources Association",  
    url = "https://aclanthology.org/2020.lrec-1.169",  
    pages = "1353--1361",  
     language = "English",  
    ISBN = "979-10-95546-34-4",}

BiSECT

@inproceedings{bisect2021,  
  title={BiSECT: Learning to Split and Rephrase Sentences with Bitexts},  
  author={Kim, Joongwon and Maddela, Mounica and Kriz, Reno and Xu, Wei and Callison-Burch, Chris},  
  booktitle={Proceedings of the Conference on Empirical Methods in Natural Language Processing (EMNLP)},  
  year={2021}}

CLEAR

@inproceedings{grabar-cardon-2018-clear,
    title = "{CLEAR} {--} Simple Corpus for Medical {F}rench",
    author = "Grabar, Natalia  and  Cardon, R{\'e}mi",
    editor = {J{\"o}nsson, Arne  and  Rennes, Evelina  and Saggion, Horacio  and  Stajner, Sanja  and  Yaneva, Victoria},
    booktitle = "Proceedings of the 1st Workshop on Automatic Text Adaptation ({ATA})",
    month = nov,
    year = "2018",
    address = "Tilburg, the Netherlands",
    publisher = "Association for Computational Linguistics",
    url = "https://aclanthology.org/W18-7002",
    doi = "10.18653/v1/W18-7002",
    pages = "3--9",
}

Wikilarge

@inproceedings{cardon-grabar-2020-french,
    title = "{F}rench Biomedical Text Simplification: When Small and Precise Helps",
    author = "Cardon, R{\'e}mi  and  Grabar, Natalia",
    editor = "Scott, Donia  and   Bel, Nuria  and   Zong, Chengqing",
    booktitle = "Proceedings of the 28th International Conference on Computational Linguistics",
    month = dec,
    year = "2020",
    address = "Barcelona, Spain (Online)",
    publisher = "International Committee on Computational Linguistics",
    url = "https://aclanthology.org/2020.coling-main.62",
    doi = "10.18653/v1/2020.coling-main.62",
    pages = "710--716",
}

frenchSIMPLIFICATION

@misc{frenchSIMPLIFICATION_2025,
    author       = { {BOURDOIS, Loïck} },  
    organization  = { {Centre Aquitain des Technologies de l'Information et Electroniques} },  
    year         = 2025,
    url          = { https://huggingface.co/datasets/CATIE-AQ/frenchSIMPLIFICATION },
    doi          = { 10.57967/hf/7134 },
    publisher    = { Hugging Face }
}

License

cc-by-4.0