Datasets:
thibaudltn
commited on
Commit
•
052f01d
1
Parent(s):
a27b560
Update README.md
Browse files
README.md
CHANGED
@@ -34,8 +34,16 @@ The version shared here is the version proposed by (Ravgofel et al., 2020) which
|
|
34 |
|
35 |
The dataset is divided between train (257,000 samples), test (99,000 samples) and dev (40,000 samples) sets.
|
36 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
37 |
Below are presented the classifiaction and sensitive attribtues labels and their proportion. Distributions are similar through the three sets.
|
38 |
|
|
|
39 |
#### Classification labels
|
40 |
|
41 |
| Profession | Numerical label | Proportion (%)| | Profession | Numerical label | Proportion (%)|
|
|
|
34 |
|
35 |
The dataset is divided between train (257,000 samples), test (99,000 samples) and dev (40,000 samples) sets.
|
36 |
|
37 |
+
To load each all splits ('train', 'dev', 'test'), use the following code :
|
38 |
+
```python
|
39 |
+
train_dataset = load_dataset("LabHC/bias_in_bios", split='train')
|
40 |
+
test_dataset = load_dataset("LabHC/bias_in_bios", split='test')
|
41 |
+
dev_dataset = load_dataset("LabHC/bias_in_bios", split='dev')
|
42 |
+
```
|
43 |
+
|
44 |
Below are presented the classifiaction and sensitive attribtues labels and their proportion. Distributions are similar through the three sets.
|
45 |
|
46 |
+
|
47 |
#### Classification labels
|
48 |
|
49 |
| Profession | Numerical label | Proportion (%)| | Profession | Numerical label | Proportion (%)|
|