metadata

configs:
  - config_name: default
    data_files:
      - split: test
        path: data/test-*
      - split: train
        path: data/train-*
      - split: validation
        path: data/validation-*
dataset_info:
  features:
    - name: tokens
      sequence: string
    - name: ner_tags
      sequence: int64
    - name: dataset
      dtype: string
  splits:
    - name: test
      num_bytes: 16147720
      num_examples: 42144
    - name: train
      num_bytes: 161576681
      num_examples: 349195
    - name: validation
      num_bytes: 12398792
      num_examples: 33464
  download_size: 43074463
  dataset_size: 190123193
task_categories:
  - token-classification
language:
  - fr
size_categories:
  - 100K<n<1M
license: cc-by-4.0

Dataset information

Dataset concatenating NER datasets, available in French and open-source, for 3 entities (LOC, PER, ORG).
There are a total of 420,264 rows, of which 346,071 are for training, 32,951 for validation and 41,242 for testing.
Our methodology is described in a blog post available in English or French.

Usage

from datasets import load_dataset
dataset = load_dataset("CATIE-AQ/frenchNER_3entities")

Dataset

Details of rows

Dataset Original	Splits	Note
Multiconer	16,548 train / 857 validation / 0 test	In practice, we use the original validation set as test set and creat a new val set from 5% of train created, i.e. 15,721 train / 827 validation / 857 test
Multinerd	140,880 train / 17,610 val / 17,695 test
Pii-masking-200k	61,958 train / 0 validation / 0 test	Only dataset without duplicate data or leaks
Wikiann	20,000 train / 10,000 val / 10,000 test
Wikiner	120,682 train / 0 validation / 13,410 test	In practice, 5% of val created from train set, i.e. 113,296 train / 5,994 validation / 13,393 test

Removing duplicate data and leaks

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

DatasetDict({
    train: Dataset({
        features: ['tokens', 'ner_tags', 'dataset'],
        num_rows: 351855
    })
    validation: Dataset({
        features: ['tokens', 'ner_tags', 'dataset'],
        num_rows: 34431
    })
    test: Dataset({
        features: ['tokens', 'ner_tags', 'dataset'],
        num_rows: 41945
    })
})

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: ['tokens', 'ner_tags', 'dataset'],
        num_rows: 346071
    })
    validation: Dataset({
        features: ['tokens', 'ner_tags', 'dataset'],
        num_rows: 32951
    })
    test: Dataset({
        features: ['tokens', 'ner_tags', 'dataset'],
        num_rows: 41242
    })
})

Note: in practice, the test split contains 8 lines which we failed to deduplicate, i.e. 0.019%.

Details of entities (after cleaning)

Datasets	Splits	O	PER	LOC	ORG
Multiconer	train	200,093	18,060	7,165	6,967
	validation	10,900	1,069	389	328
	test	11,287	979	387	381
Multinerd	train	3,041,998	149,128	105,531	68,796
	validation	410,934	17,479	13,988	3,475
	test	417,886	18,567	14,083	3,636
Pii-masking-200k	train	2,405,215	29,838	42,154	12,310
Wikiann	train	60,165	20,288	17,033	24,429
	validation	30,046	10,098	8,698	12,819
	test	31,488	10,764	9,512	13,480
Wikiner	train	2,691,294	110,079	131,839	38,988
	validation	140,935	5,481	7,204	2,121
	test	313,210	13,324	15,213	3,894
Total	train	8,398,765	327,393	303,722	151,490
	validation	592,815	34,127	30,279	18,743
	test	773,871	43,634	39,195	21,391

Columns

dataset_train = dataset['train'].to_pandas()
dataset_train.head()

     tokens 	                                            ner_tags 	                                        dataset
0 	[On, a, souvent, voulu, faire, de, La, Bruyère... 	[0, 0, 0, 0, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 0, ... 	wikiner
1 	[Les, améliorations, apportées, par, rapport, ... 	[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 2, 2, ... 	wikiner
2 	[Cette, assemblée, de, notables, ,, réunie, en... 	[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 0, ... 	wikiner
3 	[Wittgenstein, projetait, en, effet, d', élabo... 	[1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, ... 	wikiner
4 	[Le, premier, écrivain, à, écrire, des, fictio... 	[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, ... 	wikiner

the tokens column contains the tokens
the ner_tags column contains the NER tags (IOB format with 0="O", 1="PER", 2="ORG" and 3="LOC")
the dataset column identifies the row's original dataset (if you wish to apply filters to it)

Split

train corresponds to the concatenation of multiconer + multinerd + pii-masking-200k + wikiann + wikiner
validation corresponds to the concatenation of multiconer + multinerd + wikiann + wikiner
test corresponds to the concatenation of multiconer + multinerd + wikiann + wikiner

Citations

multiconer

@inproceedings{multiconer2-report,  
    title={{SemEval-2023 Task 2: Fine-grained Multilingual Named Entity Recognition (MultiCoNER 2)}},  
    author={Fetahu, Besnik and Kar, Sudipta and Chen, Zhiyu and Rokhlenko, Oleg and Malmasi, Shervin},  
    booktitle={Proceedings of the 17th International Workshop on Semantic Evaluation (SemEval-2023)},  
    year={2023},  
    publisher={Association for Computational Linguistics}}


@article{multiconer2-data,  
    title={{MultiCoNER v2: a Large Multilingual dataset for Fine-grained and Noisy Named Entity Recognition}},  
    author={Fetahu, Besnik and Chen, Zhiyu and Kar, Sudipta and Rokhlenko, Oleg and Malmasi, Shervin},  
    year={2023}}

multinerd

@inproceedings{tedeschi-navigli-2022-multinerd,  
    title = "{M}ulti{NERD}: A Multilingual, Multi-Genre and Fine-Grained Dataset for Named Entity Recognition (and Disambiguation)",  
    author = "Tedeschi, Simone and  Navigli, Roberto",  
    booktitle = "Findings of the Association for Computational Linguistics: NAACL 2022",  
    month = jul,  
    year = "2022",  
    address = "Seattle, United States",  
    publisher = "Association for Computational Linguistics",  
    url = "https://aclanthology.org/2022.findings-naacl.60",  
    doi = "10.18653/v1/2022.findings-naacl.60",  
    pages = "801--812"}

pii-masking-200k

@misc {ai4privacy_2023,  
    author = { {ai4Privacy} },  
    title = { pii-masking-200k (Revision 1d4c0a1) },  
    year = 2023,  
    url = { https://huggingface.co/datasets/ai4privacy/pii-masking-200k },  
    doi = { 10.57967/hf/1532 },  
    publisher = { Hugging Face }}

wikiann

@inproceedings{rahimi-etal-2019-massively,  
    title = "Massively Multilingual Transfer for {NER}",  
    author = "Rahimi, Afshin and Li, Yuan and Cohn, Trevor",  
    booktitle = "Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics",  
    month = jul,  
    year = "2019",  
    address = "Florence, Italy",  
    publisher = "Association for Computational Linguistics",  
    url = "https://www.aclweb.org/anthology/P19-1015",  
    pages = "151--164"}

wikiner

 @article{NOTHMAN2013151,  
    title = {Learning multilingual named entity recognition from Wikipedia},  
    journal = {Artificial Intelligence},  
    volume = {194},  
    pages = {151-175},  
    year = {2013},  
    note = {Artificial Intelligence, Wikipedia and Semi-Structured Resources},  
    issn = {0004-3702},  
    doi = {https://doi.org/10.1016/j.artint.2012.03.006},  
    url = {https://www.sciencedirect.com/science/article/pii/S0004370212000276},  
    author = {Joel Nothman and Nicky Ringland and Will Radford and Tara Murphy and James R. Curran}}

frenchNER_3entities

@misc {frenchNER2024,  
    author       = { {BOURDOIS, Loïck} },  
    organization  = { {Centre Aquitain des Technologies de l'Information et Electroniques} },  
    title        = { frenchNER_3entities },  
    year         = 2024,  
    url          = { https://huggingface.co/CATIE-AQ/frenchNER_3entities },  
    doi          = { 10.57967/hf/1751 },  
    publisher    = { Hugging Face }  
}

License

cc-by-4.0