| Property | Value |
|---|---|
| License | Apache 2.0 |
| Framework | PyTorch, TensorFlow |
| Architecture | Transformer-align (Marian) |
| BLEU Score | 54.9 (Tatoeba) |
What is opus-mt-pl-en?
opus-mt-pl-en is a specialized machine translation model developed by Helsinki-NLP, designed specifically for translating Polish text to English. With over 105,000 downloads, it's built on the Marian framework using the transformer-align architecture and trained on the OPUS dataset. The model achieves impressive performance with a BLEU score of 54.9 and a chr-F score of 0.701 on the Tatoeba test set.
Implementation Details
The model utilizes a transformer-align architecture with normalization and SentencePiece pre-processing. It's implemented using both PyTorch and TensorFlow frameworks, making it versatile for different deployment scenarios.
- Pre-processing: Normalization + SentencePiece tokenization
- Dataset: OPUS corpus
- Architecture: Marian-based transformer-align
- Evaluation Metrics: BLEU (54.9) and chr-F (0.701)
Core Capabilities
- High-quality Polish to English translation
- Support for both PyTorch and TensorFlow frameworks
- Optimized for production deployment through Inference Endpoints
- Comprehensive evaluation data available through test sets
Frequently Asked Questions
Q: What makes this model unique?
The model stands out for its specialized focus on Polish-to-English translation, achieving a high BLEU score of 54.9. It's built on the robust Marian framework and includes comprehensive evaluation data and test sets for validation.
Q: What are the recommended use cases?
This model is ideal for applications requiring high-quality Polish to English translation, such as document translation, content localization, and multilingual NLP applications. Its support for multiple frameworks makes it suitable for both research and production environments.