Opus-mt-en-dra Open-source Translation Model - Free and Efficient English to Multiple Indian Languages Translation Supported

Opus Mt En Dra

Developed by Helsinki-NLP

This model is a multilingual translation model based on Transformer architecture, supporting translation tasks from English to Kannada, Malayalam, Tamil, and Telugu.

Machine Translation

Transformers

Supports Multiple LanguagesOpen Source License:Apache-2.0 #Dravidian language translation #Low-resource language support #SentencePiece tokenization

Downloads 15.19k

Release Time : 3/2/2022

Model Overview

An English-to-Dravidian machine translation model developed by Helsinki-NLP, employing standardized preprocessing and SentencePiece tokenization technology, supporting translation for four major South Indian languages.

Model Features

Multilingual Support

A single model supports translation from English to four Dravidian languages, reducing deployment complexity.

Standardized Preprocessing

Utilizes standardized text processing and SentencePiece tokenization (spm32k) to enhance translation quality.

Language Identifier Control

Dynamic language output specification by adding target language codes (e.g., >>kan<<) at the beginning of sentences.

Model Capabilities

English to Kannada translation

English to Malayalam translation

English to Tamil translation

English to Telugu translation

Use Cases

Cross-language Communication

Travel Information Translation

Translate English travel guides into South Indian regional languages

BLEU 4.7-13.1 (varies by target language)

Content Localization

Multilingual Content Generation

Generate localized content for South Indian regional users

Supports output in four major Dravidian languages

🚀 English - Dravidian Languages Translation Model

This project focuses on translating English to Dravidian languages. It provides a transformer - based model with specific pre - processing steps and offers various evaluation metrics on test sets.

✨ Features

Supports translation from English to multiple Dravidian languages including Kannada (kan), Malayalam (mal), Tamil (tam), and Telugu (tel).
Utilizes a transformer model with normalization and SentencePiece pre - processing.
Requires a sentence - initial language token in the form of >>id<< (where id is a valid target language ID).

📦 Installation

The original weights of the model can be downloaded from opus - 2020 - 07 - 26.zip.

📚 Documentation

Model Details

Property	Details
Model Type	Transformer
Source Language(s)	eng
Target Language(s)	kan, mal, tam, tel
Pre - processing	normalization + SentencePiece (spm32k,spm32k)
OPUS Readme	eng - dra
Download Original Weights	opus - 2020 - 07 - 26.zip
Test Set Translations	opus - 2020 - 07 - 26.test.txt
Test Set Scores	opus - 2020 - 07 - 26.eval.txt

System Info

hf_name: eng - dra
source_languages: eng
target_languages: dra
opus_readme_url: https://github.com/Helsinki-NLP/Tatoeba-Challenge/tree/master/models/eng-dra/README.md
original_repo: Tatoeba - Challenge
tags: ['translation']
languages: ['en', 'ta', 'kn', 'ml', 'te', 'dra']
src_constituents: {'eng'}
tgt_constituents: {'tam', 'kan', 'mal', 'tel'}
src_multilingual: False
tgt_multilingual: True
prepro: normalization + SentencePiece (spm32k,spm32k)
url_model: https://object.pouta.csc.fi/Tatoeba-MT-models/eng-dra/opus-2020-07-26.zip
url_test_set: https://object.pouta.csc.fi/Tatoeba-MT-models/eng-dra/opus-2020-07-26.test.txt
src_alpha3: eng
tgt_alpha3: dra
short_pair: en - dra
chrF2_score: 0.46299999999999997
bleu: 10.7
brevity_penalty: 1.0
ref_len: 7928.0
src_name: English
tgt_name: Dravidian languages
train_date: 2020 - 07 - 26
src_alpha2: en
tgt_alpha2: dra
prefer_old: False
long_pair: eng - dra
helsinki_git_sha: 480fcbe0ee1bf4774bcbe6226ad9f58e63f6c535
transformers_git_sha: 2207e5d8cb224e954a7cba69fa4ac2309e9ff30b
port_machine: brutasse
port_time: 2020 - 08 - 21 - 14:41

🔧 Technical Details

Benchmarks

Testset	BLEU	chr - F
Tatoeba - test.eng - kan.eng.kan	4.7	0.348
Tatoeba - test.eng - mal.eng.mal	13.1	0.515
Tatoeba - test.eng.multi	10.7	0.463
Tatoeba - test.eng - tam.eng.tam	9.0	0.444
Tatoeba - test.eng - tel.eng.tel	7.1	0.363

📄 License

This project 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