| Property | Value |
|---|---|
| License | Apache 2.0 |
| Developer | Helsinki-NLP |
| BLEU Score | 41.3 |
| chrF Score | 0.588 |
What is opus-mt-ko-en?
opus-mt-ko-en is a specialized machine translation model developed by Helsinki-NLP for translating Korean text to English. The model demonstrates strong performance with a BLEU score of 41.3 on the Tatoeba test set, making it suitable for production-level translation tasks.
Implementation Details
The model utilizes a transformer-align architecture and implements SentencePiece tokenization with spm32k vocabulary for both source and target languages. It supports multiple Korean script variants including Hangul, Hanja, and Latin transliterations, ensuring comprehensive coverage of Korean text inputs.
- Preprocessing: Normalization + SentencePiece (spm32k,spm32k)
- Architecture: Transformer-align
- Training Date: June 17, 2020
- Source Languages: Korean (including variants)
- Target Language: English
Core Capabilities
- High-quality Korean to English translation
- Support for multiple Korean script variants
- Optimized for Tatoeba dataset performance
- Production-ready with 27,727+ downloads
Frequently Asked Questions
Q: What makes this model unique?
This model stands out for its comprehensive handling of Korean script variants and strong performance metrics, achieving a 41.3 BLEU score and 0.588 chrF score on standardized tests. It's particularly well-suited for Korean-English translation tasks requiring high accuracy.
Q: What are the recommended use cases?
The model is recommended for production environments requiring Korean to English translation, including content localization, document translation, and automated translation systems. It's particularly effective for scenarios where accurate handling of various Korean script formats is necessary.