FastPDN
| Property | Value |
|---|---|
| Parameter Count | 124M |
| License | CC-BY-4.0 |
| Language | Polish |
| Task | Named Entity Recognition |
| Framework | PyTorch, Transformers |
What is FastPDN?
FastPDN (FastPolDeepNer) is an advanced Named Entity Recognition model specifically designed for Polish language processing. It represents an evolution from its predecessor PolDeepNer2, implementing a sophisticated pipeline that leverages hydra, PyTorch, and transformers for efficient text analysis.
Implementation Details
The model is built upon two powerful pretrained models: herbert-base-case and distiluse-base-multilingual-cased-v1. It processes Polish text to identify and classify named entities with impressive accuracy, achieving an F1 score of 0.68 and precision of 0.7 on test data.
- Built with PyTorch and Transformers framework
- Trained on kpwr and cen datasets with 82 class versions
- Implements efficient token classification strategy
- Supports both inference endpoints and local deployment
Core Capabilities
- Named Entity Recognition for Polish text
- Multi-class entity classification (82 classes)
- High-accuracy entity detection (97% accuracy on test set)
- Simple integration through Transformers pipeline
- Support for both tokenized and raw text input
Frequently Asked Questions
Q: What makes this model unique?
FastPDN's specialization in Polish language NER, combined with its extensive 82-class classification system and high accuracy metrics, makes it a standout choice for Polish text analysis. Its implementation allows for easy integration and flexible deployment options.
Q: What are the recommended use cases?
The model is ideal for Polish text processing tasks requiring named entity recognition, including information extraction, document analysis, and automated text processing systems. It's particularly effective for identifying and classifying person names, locations, and other entity types in Polish text.
