SaBERT-Spanish-Fake-News

Spanish Fake News Classifier

Overview

This BERT-based text classifier was developed as a thesis project for the Computer Engineering degree at Universidad de Buenos Aires (UBA). The model is designed to detect fake news in Spanish and was fine-tuned on the dccuchile/bert-base-spanish-wwm-uncased model using a specific set of hyperparameters. It was trained on a dataset containing 125,000 Spanish news articles collected from various regions, both true and false.

Team Members

• Azul Fuentes
• Dante Reinaudo
• Lucía Pardo
• Roberto Iskandarani

Model Details

• Base Mode: dccuchile/bert-base-spanish-wwm-uncased

• Hyperparameters:

  • dropout_rate = 0.1
  • num_classes = 2
  • max_length = 128
  • batch_size = 16
  • num_epochs = 5
  • learning_rate = 3e-5

• Dataset: 125,000 Spanish news articles (True and False)

Metrics

The model's performance was evaluated using the following metrics:

• Accuracy = 83.17%
• F1-Score = 81.94%
• Precision = 85.62%
• Recall = 81.10%

Usage

Installation

You can install the required dependencies using pip:

pip install transformers torch

Loading the Model

from transformers import BertForSequenceClassification, BertTokenizer

model = BertForSequenceClassification.from_pretrained("VerificadoProfesional/SaBERT-Spanish-Fake-News")
tokenizer = BertTokenizer.from_pretrained("VerificadoProfesional/SaBERT-Spanish-Fake-News")

Predict Function

def predict(model,tokenizer,text,threshold = 0.5):   
        inputs = tokenizer(text, return_tensors="pt", padding=True, truncation=True, max_length=512)
        with torch.no_grad():
            outputs = model(**inputs)
        
        logits = outputs.logits
        probabilities = torch.softmax(logits, dim=1).squeeze().tolist()
        
        predicted_class = torch.argmax(logits, dim=1).item()
        if probabilities[predicted_class] <= threshold and predicted_class == 1:
            predicted_class = 0
  
        return bool(predicted_class), probabilities

Making Predictions

text = "Your Spanish news text here"
predicted_label,probabilities = predict(model,tokenizer,text)
print(f"Text: {text}")
print(f"Predicted Class: {predicted_label}")
print(f"Probabilities: {probabilities}")

License

Apache License 2.0

Acknowledgments

Special thanks to DCC UChile for the base Spanish BERT model and to all contributors to the dataset used for training.