Opus Mt En Zle

🚀 eng-zle Translation Model

This project focuses on the translation from English to East Slavic languages. It provides a transformer-based model with specific pre - processing steps and offers benchmarks for evaluation.

🚀 Quick Start

The eng-zle model is designed for translating English to East Slavic languages. You can download the original weights and test set translations from the provided links.

Download Links

• Original Weights: opus2m - 2020 - 08 - 02.zip
• Test Set Translations: opus2m - 2020 - 08 - 02.test.txt
• Test Set Scores: opus2m - 2020 - 08 - 02.eval.txt

Pre - processing

The model uses normalization + SentencePiece (spm32k, spm32k) for pre - processing. A sentence initial language token in the form of >>id<< (id = valid target language ID) is required.

✨ Features

• Language Support: Supports translation from English (eng) to multiple East Slavic languages including Belarusian (bel, bel_Latn), Old East Slavic (orv_Cyrl), Rusyn (rue), Russian (rus), and Ukrainian (ukr).
• Transformer Model: Utilizes a transformer - based architecture for translation.

📚 Documentation

Model Information

System Information

• hf_name: eng - zle
• source_languages: eng
• target_languages: zle
• opus_readme_url: https://github.com/Helsinki-NLP/Tatoeba-Challenge/tree/master/models/eng-zle/README.md
• original_repo: Tatoeba - Challenge
• tags: ['translation']
• languages: ['en', 'be', 'ru', 'uk', 'zle']
• src_constituents: {'eng'}
• tgt_constituents: {'bel', 'orv_Cyrl', 'bel_Latn', 'rus', 'ukr', 'rue'}
• src_multilingual: False
• tgt_multilingual: True
• prepro: Normalization + SentencePiece (spm32k, spm32k)
• url_model: https://object.pouta.csc.fi/Tatoeba-MT-models/eng-zle/opus2m-2020-08-02.zip
• url_test_set: https://object.pouta.csc.fi/Tatoeba-MT-models/eng-zle/opus2m-2020-08-02.test.txt
• src_alpha3: eng
• tgt_alpha3: zle
• short_pair: en - zle
• chrF2_score: 0.58
• bleu: 37.2
• brevity_penalty: 0.9890000000000001
• ref_len: 63493.0
• src_name: English
• tgt_name: East Slavic languages
• src_alpha2: en
• tgt_alpha2: zle
• prefer_old: False
• long_pair: eng - zle
• helsinki_git_sha: 480fcbe0ee1bf4774bcbe6226ad9f58e63f6c535
• transformers_git_sha: 2207e5d8cb224e954a7cba69fa4ac2309e9ff30b
• port_machine: brutasse
• port_time: 2020 - 08 - 21 - 14:41

📄 License

This project is licensed under the Apache - 2.0 license.

Benchmarks

The following table shows the BLEU and chr - F scores for different test sets: