Fabio Passos
commited on
Commit
·
3402a30
1
Parent(s):
8465782
first commit
Browse files- README.md +54 -3
- models/miso_br_rf_classifier.joblib +3 -0
- models/preprocessing_config.pkl +3 -0
- requirements.txt +4 -0
README.md
CHANGED
|
@@ -1,3 +1,54 @@
|
|
| 1 |
-
|
| 2 |
-
|
| 3 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# MISO-BR Misogyny Classifier
|
| 2 |
+
|
| 3 |
+
This model classifies text in Brazilian Portuguese as misogynistic or non-misogynistic. It's trained on the [MISO-BR dataset](https://huggingface.co/datasets/fabiopassos/miso-br).
|
| 4 |
+
|
| 5 |
+
## Model Details
|
| 6 |
+
|
| 7 |
+
- **Model Type**: TF-IDF + RandomForest classifier
|
| 8 |
+
- **Language**: Portuguese (Brazil)
|
| 9 |
+
- **Task**: Binary classification (misogynistic vs non-misogynistic content)
|
| 10 |
+
- **Framework**: scikit-learn
|
| 11 |
+
|
| 12 |
+
## Performance
|
| 13 |
+
|
| 14 |
+
The model was evaluated on a test set and achieved:
|
| 15 |
+
|
| 16 |
+
- **F1 Score (macro)**: 0.6758
|
| 17 |
+
- **Accuracy**: 0.6778
|
| 18 |
+
- **AUC**: 0.7314
|
| 19 |
+
|
| 20 |
+
## Usage
|
| 21 |
+
|
| 22 |
+
```python
|
| 23 |
+
from huggingface_hub import hf_hub_download
|
| 24 |
+
import joblib
|
| 25 |
+
import spacy
|
| 26 |
+
|
| 27 |
+
# Download the model from Hugging Face Hub
|
| 28 |
+
model_path = hf_hub_download(repo_id="fabiopassos/miso-br-classifier",
|
| 29 |
+
filename="models/miso_br_rf_classifier.joblib")
|
| 30 |
+
|
| 31 |
+
# Load the model
|
| 32 |
+
model = joblib.load(model_path)
|
| 33 |
+
|
| 34 |
+
# Load spaCy for Portuguese
|
| 35 |
+
nlp = spacy.load("pt_core_news_sm")
|
| 36 |
+
|
| 37 |
+
# Preprocess function
|
| 38 |
+
def preprocess_text(text):
|
| 39 |
+
doc = nlp(text)
|
| 40 |
+
tokens = [token.lemma_.lower() for token in doc
|
| 41 |
+
if not token.is_stop and not token.is_punct and token.is_alpha]
|
| 42 |
+
return " ".join(tokens)
|
| 43 |
+
|
| 44 |
+
# Example text
|
| 45 |
+
text = "Seu texto para classificar aqui"
|
| 46 |
+
processed_text = preprocess_text(text)
|
| 47 |
+
|
| 48 |
+
# Predict
|
| 49 |
+
prediction = model.predict([processed_text])[0]
|
| 50 |
+
probability = model.predict_proba([processed_text])[0][1]
|
| 51 |
+
|
| 52 |
+
print(f"Texto: {text}")
|
| 53 |
+
print(f"É misógino: {'Sim' if prediction == 1 else 'Não'}")
|
| 54 |
+
print(f"Probabilidade: {probability:.4f}")
|
models/miso_br_rf_classifier.joblib
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:cad11eeaffb1bc4112bd47c86791a6dcd5f81a363d4ddaacc36770f5483b1095
|
| 3 |
+
size 1177125
|
models/preprocessing_config.pkl
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:0d7203b61175a2eabd286bdb156f984d12e47e2ac8bf61b8ffde9677f5b21cc4
|
| 3 |
+
size 3498
|
requirements.txt
ADDED
|
@@ -0,0 +1,4 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
scikit-learn==1.7.0
|
| 2 |
+
spacy==3.7.2
|
| 3 |
+
joblib>=1.3.0
|
| 4 |
+
pt_core_news_sm @ https://github.com/explosion/spacy-models/releases/download/pt_core_news_sm-3.7.0/pt_core_news_sm-3.7.0-py3-none-any.whl
|