| Property | Value |
|---|---|
| Base Model | BERT-base-spanish-wwm-uncased |
| Training Dataset | 50,000 Movie Reviews |
| Accuracy | 91.01% |
| F1 Score | 90.88% |
| Author | Eduardo Muñoz |
What is beto_sentiment_analysis_es?
beto_sentiment_analysis_es is a specialized sentiment analysis model built on the BETO architecture, specifically designed for Spanish language text analysis. The model has been fine-tuned using Amazon SageMaker on a comprehensive dataset of 50,000 movie reviews, making it particularly effective for sentiment classification tasks in Spanish text.
Implementation Details
The model is based on BETO, a Spanish BERT-base model trained using Whole Word Masking technique. It was fine-tuned using carefully selected hyperparameters including a learning rate of 3e-05, batch size of 32, and trained for 4 epochs. The training utilized mixed precision (FP16) for optimal performance.
- Training Set: 42,500 reviews
- Validation Set: 3,750 reviews
- Test Set: 3,750 reviews
- Precision: 91.06%
- Recall: 90.71%
Core Capabilities
- Binary sentiment classification for Spanish text
- Specialized in movie review analysis
- Adaptable to general review sentiment analysis
- High accuracy and balanced performance metrics
Frequently Asked Questions
Q: What makes this model unique?
This model combines the power of BETO's Spanish language understanding with specific fine-tuning for sentiment analysis, achieving over 91% accuracy on movie reviews while maintaining balanced precision and recall metrics.
Q: What are the recommended use cases?
While primarily optimized for movie review sentiment analysis, the model can be effectively applied to various Spanish text classification tasks, including product reviews, social media sentiment analysis, and general opinion mining in Spanish content.