opus-mt-en-tut Open-source Translation Model - Free Translation from English to Multiple Altaic Languages

Opus Mt En Tut

Developed by Helsinki-NLP

This is a multilingual machine translation model based on the Transformer architecture, supporting translation tasks from English to various Altaic languages.

Machine Translation

Transformers

Supports Multiple LanguagesOpen Source License:Apache-2.0 #Multilingual Altaic Translation #Low-resource Language Support #SentencePiece Tokenization

Downloads 38

Release Time : 3/2/2022

Model Overview

This model focuses on translation tasks from English to Altaic languages, supporting multiple languages including Turkish, Kazakh, Mongolian, etc. It employs standardized preprocessing and SentencePiece tokenization technology, requiring the addition of target language identifiers at the beginning of sentences for translation.

Model Features

Multilingual Support

Supports translation tasks from English to 20 Altaic languages

Preprocessing Optimization

Utilizes standardized preprocessing and SentencePiece tokenization technology (spm32k)

Target Language Identification

Requires adding target language identifiers (e.g., >>tur<<) at the beginning of sentences for translation

Model Capabilities

English to Altaic language translation

Multilingual machine translation

Supports multiple writing systems (Latin/Cyrillic/Arabic)

Use Cases

Cross-language Communication

English-Turkish Translation

Translate English content into Turkish

Test set BLEU 34.2

English-Kazakh Translation

Translate English content into Kazakh

Test set BLEU 11.8

Academic Research

Altaic Language Studies

Assist linguists in comparative studies of Altaic languages

🚀 eng-tut

This is a translation model that translates from English to Altaic languages. It provides a range of features and performance metrics for language translation tasks.

✨ Features

Language Pair: Translates from English (eng) to multiple Altaic languages including aze_Latn, bak, chv, etc.
Model Type: Uses a transformer model.
Pre - processing: Applies normalization and SentencePiece (spm32k, spm32k) for pre - processing.
Language Token Requirement: A sentence initial language token in the form of >>id<< (id = valid target language ID) is required.

📦 Installation

There is no specific installation command provided in the original README. So, this section is skipped.

📚 Documentation

Model Information

Property	Details
Model Type	transformer
Source Language(s)	eng
Target Language(s)	aze_Latn, bak, chv, crh, crh_Latn, kaz_Cyrl, kaz_Latn, kir_Cyrl, kjh, kum, mon, nog, ota_Arab, ota_Latn, sah, tat, tat_Arab, tat_Latn, tuk, tuk_Latn, tur, tyv, uig_Arab, uig_Cyrl, uzb_Cyrl, uzb_Latn, xal
Pre - processing	normalization + SentencePiece (spm32k, spm32k)
Download 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

System Info

hf_name: eng - tut
source_languages: eng
target_languages: tut
opus_readme_url: https://github.com/Helsinki-NLP/Tatoeba-Challenge/tree/master/models/eng-tut/README.md
original_repo: Tatoeba - Challenge
tags: ['translation']
languages: ['en', 'tut']
src_constituents: {'eng'}
tgt_constituents: set()
src_multilingual: False
tgt_multilingual: True
prepro: normalization + SentencePiece (spm32k, spm32k)
url_model: https://object.pouta.csc.fi/Tatoeba-MT-models/eng-tut/opus2m-2020-08-02.zip
url_test_set: https://object.pouta.csc.fi/Tatoeba-MT-models/eng-tut/opus2m-2020-08-02.test.txt
src_alpha3: eng
tgt_alpha3: tut
short_pair: en - tut
chrF2_score: 0.451
bleu: 19.5
brevity_penalty: 1.0
ref_len: 57472.0
src_name: English
tgt_name: Altaic languages
train_date: 2020 - 08 - 02
src_alpha2: en
tgt_alpha2: tut
prefer_old: False
long_pair: eng - tut
helsinki_git_sha: 480fcbe0ee1bf4774bcbe6226ad9f58e63f6c535
transformers_git_sha: 2207e5d8cb224e954a7cba69fa4ac2309e9ff30b
port_machine: brutasse
port_time: 2020 - 08 - 21 - 14:41

Benchmarks

testset	BLEU	chr - F
newsdev2016 - entr - engtur.eng.tur	10.4	0.438
newstest2016 - entr - engtur.eng.tur	9.1	0.414
newstest2017 - entr - engtur.eng.tur	9.5	0.414
newstest2018 - entr - engtur.eng.tur	9.5	0.415
Tatoeba - test.eng - aze.eng.aze	27.2	0.580
Tatoeba - test.eng - bak.eng.bak	5.8	0.298
Tatoeba - test.eng - chv.eng.chv	4.6	0.301
Tatoeba - test.eng - crh.eng.crh	6.5	0.342
Tatoeba - test.eng - kaz.eng.kaz	11.8	0.360
Tatoeba - test.eng - kir.eng.kir	24.6	0.499
Tatoeba - test.eng - kjh.eng.kjh	2.2	0.052
Tatoeba - test.eng - kum.eng.kum	8.0	0.229
Tatoeba - test.eng - mon.eng.mon	10.3	0.362
Tatoeba - test.eng.multi	19.5	0.451
Tatoeba - test.eng - nog.eng.nog	1.5	0.117
Tatoeba - test.eng - ota.eng.ota	0.2	0.035
Tatoeba - test.eng - sah.eng.sah	0.7	0.080
Tatoeba - test.eng - tat.eng.tat	10.8	0.320
Tatoeba - test.eng - tuk.eng.tuk	5.6	0.323
Tatoeba - test.eng - tur.eng.tur	34.2	0.623
Tatoeba - test.eng - tyv.eng.tyv	8.1	0.192
Tatoeba - test.eng - uig.eng.uig	0.1	0.158
Tatoeba - test.eng - uzb.eng.uzb	4.2	0.298
Tatoeba - test.eng - xal.eng.xal	0.1	0.061

📄 License

The model is licensed under the apache - 2.0 license.

Featured Recommended AI Models

Empowering the Future, Your AI Solution Knowledge Base

English 简体中文繁體中文にほんご