Downloads · 30 days
85
0% of all-time downloads
french_toxicity_classifier_plus_v2
EIStakovskii/french_toxicity_classifier_plus_v2
french_toxicity_classifier_plus_v2 is a text classification model from EIStakovskii. Use it when you need a label for a piece of text. It is set up for transformers. The card lists the license as other.
NB: this version of the model is the improved version of EIStakovskii/frenchtoxicityclassifierplus. To see the source code of training and the data please follow the github link.
All-time downloads
41.4K
Public
Parameters
111M
886 MB on disk
Likes
4
Public
Repo makeup
Click a slice to open those files.
.bin443 MB · 50%
Parameter types
How the weights are stored.
F32111M · 100%
Try a prompt
Model card
From the Hugging Face model README
Description
NB: this version of the model is the improved version of EIStakovskii/french_toxicity_classifier_plus. To see the source code of training and the data please follow the github link.
This model was trained for toxicity labeling.
The model was fine-tuned based off the CamemBERT language model.
To use the model:
from transformers import pipeline
classifier = pipeline("text-classification", model = 'EIStakovskii/french_toxicity_classifier_plus_v2')
print(classifier("Foutez le camp d'ici!"))
Metrics (at validation):
| epoch | step | eval_accuracy | eval_f1 | eval_loss |
|---|---|---|---|---|
| 1.16 | 1600 | 0.9015412511332729 | 0.8968269048071442 | 0.3014959990978241 |
Comparison against Perspective
This model was compared against the Google's Perspective API that similarly detects toxicity. Two models were tested on two datasets: the size of 200 sentences and 400 sentences. The first one (arguably harder) was collected from the sentences of the JigSaw and DeTox datasets. The second one (easier) was collected from the combination of sources: both from JigSaw and DeTox as well as Paradetox translations and sentences extracted from Reverso Context by keywords.
french_toxicity_classifier_plus_v2
| size | accuracy | f1 |
|---|---|---|
| 200 | 0.783 | 0.803 |
| 400 | 0.890 | 0.879 |
Perspective
| size | accuracy | f1 |
|---|---|---|
| 200 | 0.826 | 0.795 |
| **400 | 0.632 | 0.418 |
**I suspect that Perspective has such a low score in the case of the FR dataset (400) because it refuses to trigger on the words "merde" and "putain" and some more rarer words in French like "cul" and so on.
- Task
- Text Classification
- Library
- transformers
- Type
- camembert
- License
- other
- Languages
- fr
- Created
- Dec 2, 2022
- Updated
- Mar 24, 2023