pun-recognition-pt
| Property | Value |
|---|---|
| Author | Superar |
| Task | Text Classification (Pun Recognition) |
| Language | Portuguese |
| Performance | 69% F1-Score |
| Paper | LREC-COLING 2024 |
What is pun-recognition-pt?
pun-recognition-pt is a specialized natural language processing model designed for detecting puns in Portuguese text. It was developed as part of research presented at PROPOR 2024 and LREC-COLING 2024, utilizing the Puntuguese dataset, which consists of carefully curated puns and corresponding non-pun texts in Portuguese.
Implementation Details
The model employs a fine-tuned transformer architecture optimized through extensive hyperparameter tuning using Weights and Biases. The optimal configuration was achieved with AdamW optimizer, learning rate of 8.47e-5, batch size of 128, and 2 training epochs.
- Built on transformer architecture for text classification
- Optimized through random search hyperparameter tuning
- Integrates easily with Hugging Face's transformers library
- Supports GPU acceleration for inference
Core Capabilities
- Accurate pun detection in Portuguese texts
- Binary classification of text as pun or non-pun
- Batch processing support
- Easy integration with popular ML frameworks
Frequently Asked Questions
Q: What makes this model unique?
This model is specifically trained for Portuguese pun recognition, a challenging task in computational humor detection. It's one of the first models dedicated to this task for the Portuguese language, achieving significant performance metrics.
Q: What are the recommended use cases?
The model is ideal for computational linguistics research, humor detection systems, content analysis tools, and educational applications focusing on Portuguese language wordplay. It can be integrated into larger NLP pipelines for humor analysis.
