| Property | Value |
|---|---|
| Model Type | Transformer-align |
| Developer | Helsinki-NLP |
| Release Date | June 17, 2020 |
| BLEU Score | 27.5 |
| chrF Score | 0.479 |
| Model URL | https://huggingface.co/Helsinki-NLP/opus-mt-pl-no |
What is opus-mt-pl-no?
opus-mt-pl-no is a specialized neural machine translation model designed to translate text from Polish to Norwegian. Developed by Helsinki-NLP, this model utilizes the transformer-align architecture and implements SentencePiece tokenization with 4k vocabulary size for both source and target languages.
Implementation Details
The model employs a transformer-align architecture with normalization and SentencePiece preprocessing. It was trained on the OPUS dataset and demonstrates strong performance with a BLEU score of 27.5 and a chrF score of 0.479 on the Tatoeba test set.
- Pre-processing: Normalization + SentencePiece (spm4k,spm4k)
- Source Language Support: Polish (pl)
- Target Language Support: Norwegian (no), including both Bokmål (nob) and Nynorsk (nno)
- Training Date: June 17, 2020
Core Capabilities
- High-quality Polish to Norwegian translation
- Support for both Norwegian written standards (Bokmål and Nynorsk)
- Optimized for general-purpose translation tasks
- Demonstrated reliability with a strong BLEU score of 27.5
Frequently Asked Questions
Q: What makes this model unique?
This model specializes in Polish to Norwegian translation, supporting both major Norwegian written standards. Its transformer-align architecture and carefully tuned tokenization make it particularly effective for this language pair.
Q: What are the recommended use cases?
The model is well-suited for general-purpose translation tasks between Polish and Norwegian, including document translation, web content localization, and automated translation services requiring Polish to Norwegian conversion.