Update README.md
Browse files
README.md
CHANGED
@@ -2,36 +2,26 @@
|
|
2 |
language: br
|
3 |
task_categories:
|
4 |
- token-classification
|
5 |
-
dataset_info:
|
6 |
-
features:
|
7 |
-
- name: tokens
|
8 |
-
sequence: string
|
9 |
-
- name: ner_tags
|
10 |
-
sequence: int64
|
11 |
-
splits:
|
12 |
-
- name: train
|
13 |
-
num_bytes: 127019
|
14 |
-
num_examples: 915
|
15 |
-
- name: validation
|
16 |
-
num_bytes: 121393
|
17 |
-
num_examples: 946
|
18 |
-
- name: test
|
19 |
-
num_bytes: 130972
|
20 |
-
num_examples: 952
|
21 |
-
download_size: 122664
|
22 |
-
dataset_size: 379384
|
23 |
configs:
|
24 |
-
- config_name:
|
25 |
data_files:
|
26 |
- split: train
|
27 |
-
path: data/train-*
|
28 |
- split: validation
|
29 |
-
path: data/validation-*
|
30 |
- split: test
|
31 |
-
path: data/test-*
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
32 |
---
|
33 |
|
34 |
-
|
35 |
Version nettoyée de [WikiAnn](https://huggingface.co/datasets/tner/wikiann).
|
36 |
En effet, la version originale contenait des leaks et des duplications.
|
37 |
De 1000 effectifs par split, la nouvelle répartition devient alors la suivante :
|
@@ -53,8 +43,10 @@ DatasetDict({
|
|
53 |
})
|
54 |
```
|
55 |
|
|
|
|
|
56 |
|
57 |
-
###
|
58 |
Le dictionnaire label2id est disponible [ici](https://huggingface.co/datasets/tner/wikiann/raw/main/dataset/label.json).
|
59 |
```python
|
60 |
{
|
@@ -68,10 +60,31 @@ Le dictionnaire label2id est disponible [ici](https://huggingface.co/datasets/tn
|
|
68 |
}
|
69 |
```
|
70 |
|
71 |
-
|
72 |
|
73 |
| | O | I-ORG | B-LOC | B-ORG | B-PER | I-LOC | I-PER |
|
74 |
| ---------- | ---- | ----- | ----- | ----- | ----- | ----- | ----- |
|
75 |
| train | 4502 | 525 | 360 | 710 | 284 | 592 | 280 |
|
76 |
| validation | 4179 | 427 | 396 | 767 | 283 | 581 | 292 |
|
77 |
-
| test | 4491 | 505 | 336 | 639 | 348 | 808 | 345 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2 |
language: br
|
3 |
task_categories:
|
4 |
- token-classification
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
5 |
configs:
|
6 |
+
- config_name: avec_prefixe
|
7 |
data_files:
|
8 |
- split: train
|
9 |
+
path: data/avec_prefixe/train-*
|
10 |
- split: validation
|
11 |
+
path: data/avec_prefixe/validation-*
|
12 |
- split: test
|
13 |
+
path: data/avec_prefixe/test-*
|
14 |
+
- config_name: sans_prefixe
|
15 |
+
data_files:
|
16 |
+
- split: train
|
17 |
+
path: data/sans_prefixe/train-*
|
18 |
+
- split: validation
|
19 |
+
path: data/sans_prefixe/validation-*
|
20 |
+
- split: test
|
21 |
+
path: data/sans_prefixe/test-*
|
22 |
---
|
23 |
|
24 |
+
# Description
|
25 |
Version nettoyée de [WikiAnn](https://huggingface.co/datasets/tner/wikiann).
|
26 |
En effet, la version originale contenait des leaks et des duplications.
|
27 |
De 1000 effectifs par split, la nouvelle répartition devient alors la suivante :
|
|
|
43 |
})
|
44 |
```
|
45 |
|
46 |
+
# Configurations
|
47 |
+
Le jeu de données est disponible en deux configurations avec prefixe (i.e. schéma [IOB](https://en.wikipedia.org/wiki/Inside%E2%80%93outside%E2%80%93beginning_(tagging))) et sans préfixe.
|
48 |
|
49 |
+
### Avec prefixe
|
50 |
Le dictionnaire label2id est disponible [ici](https://huggingface.co/datasets/tner/wikiann/raw/main/dataset/label.json).
|
51 |
```python
|
52 |
{
|
|
|
60 |
}
|
61 |
```
|
62 |
|
63 |
+
La répartition des entités est alors la suivante
|
64 |
|
65 |
| | O | I-ORG | B-LOC | B-ORG | B-PER | I-LOC | I-PER |
|
66 |
| ---------- | ---- | ----- | ----- | ----- | ----- | ----- | ----- |
|
67 |
| train | 4502 | 525 | 360 | 710 | 284 | 592 | 280 |
|
68 |
| validation | 4179 | 427 | 396 | 767 | 283 | 581 | 292 |
|
69 |
+
| test | 4491 | 505 | 336 | 639 | 348 | 808 | 345 |
|
70 |
+
|
71 |
+
|
72 |
+
### Sans prefixe
|
73 |
+
Le dictionnaire label2id est alors
|
74 |
+
```python
|
75 |
+
{
|
76 |
+
"O": 0,
|
77 |
+
"PER": 1,
|
78 |
+
"ORG": 2,
|
79 |
+
"LOC": 3
|
80 |
+
}
|
81 |
+
|
82 |
+
```
|
83 |
+
|
84 |
+
Et la répartition des entités devient
|
85 |
+
|
86 |
+
| | O | LOC | PER | ORG |
|
87 |
+
| ---------- | ---- | --- | ---- | ---- |
|
88 |
+
| train | 4502 | 805 | 1070 | 876 |
|
89 |
+
| validation | 4179 | 719 | 1163 | 864 |
|
90 |
+
| test | 4491 | 850 | 975 | 1156 |
|