Update README.md

README.md

@@ -39,7 +39,7 @@ license: cc-by-4.0
 
 # Dataset information 
 Dataset concatenating all NER datasets, available in French and open-source, for 3 entities (LOC, PER, ORG).  
-There are a total of **424,803** rows, of which 349,195 are for training, 33,464 for validation and 42,144 for testing.  
+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](https://blog.vaniila.ai/en/NER_en/) or [French](https://blog.vaniila.ai/NER/).
 
 
@@ -48,85 +48,62 @@ Our methodology is described in a blog post available in [English](https://blog.
 from datasets import load_dataset
 dataset = load_dataset("CATIE-AQ/frenchNER")
 ```
+
+
+# Dataset
+## Details of rows
+| Dataset Original    | Splits  | Note  |
+| ----------- | ----------- | ----------- |
+| [Multiconer](https://huggingface.co/datasets/aashsach/multiconer2)| 16,548 train / 857 validation / 0 test  | In practice, we use the original validation set as test set<br> and creat a new val set from 5% of train created, i.e.<br> 15,721 train / 827 validation / 857 test|       
+| [Multinerd](https://huggingface.co/datasets/Babelscape/multinerd)| 140,880 train /  17,610 val / 17,695 test |   |       
+| [Pii-masking-200k](https://huggingface.co/datasets/ai4privacy/pii-masking-200k)| 61,958 train / 0 validation / 0 test  | Only dataset without duplicate data or leaks |       
+| [Wikiann](https://huggingface.co/datasets/wikiann)| 20,000 train / 10,000 val / 10,000 test  |   |       
+| [Wikiner](https://huggingface.co/datasets/Jean-Baptiste/wikiner_fr)| 120,682 train / 0 validation / 13,410 test | In practice, 5% of val created from train set, i.e.<br> 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: ['text', 'summary', 'dataset'],
+        num_rows: 351855
+    })
+    validation: Dataset({
+        features: ['text', 'summary', 'dataset'],
+        num_rows: 34431
+    })
+    test: Dataset({
+        features: ['text', 'summary', '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: 349195
+        num_rows: 346071
     })
     validation: Dataset({
         features: ['tokens', 'ner_tags', 'dataset'],
-        num_rows: 33464
+        num_rows: 32951
     })
     test: Dataset({
         features: ['tokens', 'ner_tags', 'dataset'],
-        num_rows: 42144
+        num_rows: 41242
     })
 })
 ```
 
 
-# Dataset
-## Dataset details
-PARLER DE LA DEDUPLICATION DES DONNEES ET DES LEAKS (INDIVIDUEL PUIS GLOBAL)
-
-### Détails lignes
-| Dataset Original    | Valeurs annoncées | Dataset Clean  |  Valeurs après Clean | Note  |
-| ----------- | ----------- | ----------- | ----------- | ----------- |
-| [Mapa](https://huggingface.co/datasets/joelniklaus/mapa)| X train / X validation / X test     | TODO | 1,259 train / 97 validation / 487 test   | X  |           
-| [Multiconer](https://huggingface.co/datasets/aashsach/multiconer2)| X train / X validation / X test      | TODO | 15,538 train / 827 validation / 855 test   | X  |       
-| [Multinerd](https://huggingface.co/datasets/Babelscape/multinerd)| X train / X validation / X test      | TODO | 137,917 train / 17,306 validation / 17,637 test   | X  |       
-| [Pii-masking-200k](https://huggingface.co/datasets/ai4privacy/pii-masking-200k)| 61,958 train / 0 validation / 0 test      | TODO | 61,958 train / 0 validation / 0 test   | No leak or duplicated data  |       
-| [Redfm](https://huggingface.co/datasets/Babelscape/REDFM)| X train / X validation / X test      | TODO | 1,865 train / 416 validation / 415 test   | X  |       
-| [Wikiann](https://huggingface.co/datasets/wikiann)| X train / X validation / X test     | TODO | 17,362 train / 8,824 validation / 9,357 test   | X  |       
-| [Wikiner](https://huggingface.co/datasets/Jean-Baptiste/wikiner_fr)| 120,682 train / 0 validation / 13,410 test     | TODO | 120,063 train / 0 validation / 13,393 test  | En pratique, 5% de val créée, donc 113,296 train / 5,994 validation / 13,393 test |
-
-Total : 
-288,309 train / 34,078 validation / 42,144 test
-287,237 train / 33,464 validation / 42,144 test
-Leaks et duplications (apparues après la concaténation : une donnée du split d'entraînement A peut ne pas être dans le split de test de A mais être présente dans le jeu de test de B et donc ça crée un leak dans le jeu de données A+B) :
-leaks in train split: 1071
-leaks in valsplit: 613
-duplicate sentences in train dataset: 1839
-duplicate sentences in val dataset: 55
-duplicate sentences in test dataset: 8
-
-Mapa
-1297 train / 97 val / 490 test
-APRES NETTOYAGE A FAIRE
-
-Multiconer
-16 548 train / 857 validation
-16 364 train / 855 validation
-leaks in train split (w.r.t val split): 13
-duplicate sentences in train dataset:  186
-duplicate sentences in validation dataset:  2
-
-MULITNERD
-140 880 train /  17 610 val / 17 695 test
-138 221 train /  17 409 val / 17 637 test
-leaks in train split: 69
-leaks in val split: 20
-duplicate sentences in train dataset: 2600
-duplicate sentences in val dataset: 201
-duplicate sentences in test dataset: 58
-
-wikiann :
-20 000 train / 10 000 val / 10 000 test
-17 370 train / 9 300 val / 9 375 test
-leaks in train split: 742
-leaks in val split: 473
-duplicate sentences in train dataset: 1889
-duplicate sentences in val dataset: 700
-duplicate sentences in test dataset: 644
-
-
-wikiner:
-leaks in train split: 23
-duplicate sentences in train dataset: 599
-duplicate sentences in test dataset: 17
-
-### Détails entitées (après nettoyage)
+### Details of entities (after cleaning)
 
 <table>
 <thead>
@@ -140,28 +117,6 @@ duplicate sentences in test dataset: 17
   </tr>
 </thead>
 <tbody>
-    <tr>
-    <td rowspan="3"><br>Mapa</td>
-    <td><br>train</td>
-    <td><br>61,959</td>
-    <td><br>745</td>
-    <td><br>208</td>
-    <td><br>314</td>
-  </tr>
-  <tr>
-    <td><br>validation</td>
-    <td><br>7,826</td>
-    <td><br>51</td>
-    <td><br>24</td>
-    <td><br>78</td>
-  </tr>
-  <tr>
-    <td><br>test</td>
-    <td><br>21,981</td>
-    <td><br>121</td>
-    <td><br>32</td>
-    <td><br>298</td>
-  </tr>
     <tr>
     <td rowspan="3"><br>Multiconer</td>
     <td><br>train</td>
@@ -213,28 +168,6 @@ duplicate sentences in test dataset: 17
     <td><br>29,838</td>
     <td><br>42,154</td>
     <td><br>12,310</td>
-  </tr>
-    <tr>
-    <td rowspan="3"><br>Redfm</td>
-    <td><br>train</td>
-    <td><br>130,152</td>
-    <td><br>2,833</td>
-    <td><br>7,889</td>
-    <td><br>4,096</td>
-  </tr>
-  <tr>
-    <td><br>validation</td>
-    <td><br>23,133</td>
-    <td><br>859</td>
-    <td><br>757</td>
-    <td><br>729</td>
-  </tr>
-  <tr>
-    <td><br>test</td>
-    <td><br>22,951</td>
-    <td><br>675</td>
-    <td><br>930</td>
-    <td><br>708</td>
   </tr>
     <tr>
     <td rowspan="3"><br>Wikiann</td>
@@ -283,24 +216,24 @@ duplicate sentences in test dataset: 17
   <tr>
     <td rowspan="3"><br>Total</td>
     <td><br>train</td>
-    <td><br><b>8,590,876</b></td>
-    <td><br><b>330,971</b></td>
-    <td><br><b>311,819</b></td>
-    <td><br><b>155,900</b></td>
+    <td><br><b>8,398,765</b></td>
+    <td><br><b>327,393</b></td>
+    <td><br><b>303,722</b></td>
+    <td><br><b>151,490</b></td>
   </tr>
   <tr>
     <td><br>validation</td>
-    <td><br><b>623,774</b></td>
-    <td><br><b>35,037</b></td>
-    <td><br><b>31,060</b></td>
-    <td><br><b>19,550</b></td>
+    <td><br><b>592,815</b></td>
+    <td><br><b>34,127</b></td>
+    <td><br><b>30,279</b></td>
+    <td><br><b>18,743</b></td>
   </tr>
   <tr>
     <td><br>test</td>
-    <td><br><b>818,803</b></td>
-    <td><br><b>44,430</b></td>
-    <td><br><b>40,157</b></td>
-    <td><br><b>22,397</b></td>
+    <td><br><b>773,871</b></td>
+    <td><br><b>43,634</b></td>
+    <td><br><b>39,195</b></td>
+    <td><br><b>21,391</b></td>
   </tr>
 </tbody>
 </table>
@@ -326,13 +259,75 @@ dataset_train.head()
 
 
 ## Split
-- `train` corresponds to the concatenation of
-- `validation` corresponds to the concatenation of
-- `test` corresponds to the concatenation of 
+- `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
+
+### 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}}
 
 
-# Citation
+### frenchNER
 ```
 A GENERER
 ```