SSE-TUC-mt-el-en-cased Open-source Greek-English Translation Model

SSE TUC Mt El En Cased

Developed by lighteternal

A Greek-English neural machine translation model jointly developed by the Hellenic Military Academy and the Technical University of Crete, based on the Transformer architecture, supporting mixed-case text processing.

Machine Translation

Transformers

Supports Multiple LanguagesOpen Source License:Apache-2.0 #Greek-English Translation #High BLEU Score #Mixed Case Handling

Downloads 15

Release Time : 3/2/2022

Model Overview

This model focuses on high-quality machine translation tasks from Greek to English, utilizing advanced Transformer architecture and BPE tokenization technology, suitable for various text translation scenarios.

Model Features

Mixed Case Support

The model can process mixed-case input text, preserving the original text's case characteristics.

BPE Tokenization

Utilizes 20k-encoded BPE (Byte Pair Encoding) tokenization technology to effectively handle out-of-vocabulary and rare words.

High-Performance Translation

Achieves a 79.3 BLEU score on the Tatoeba test set, demonstrating excellent performance.

Model Capabilities

Greek to English text translation

Mixed-case text processing

Technical term translation

Use Cases

Academic Research

Greek Academic Literature Translation

Translating Greek academic papers and research findings into English

High-quality translation results, maintaining accuracy of academic terminology

Business Applications

Business Document Translation

Translating Greek business contracts, reports, and other documents

Preserves document formatting and accuracy of professional terminology

🚀 Greek to English NMT

This project, developed by the Hellenic Army Academy (SSE) and the Technical University of Crete (TUC), focuses on Greek to English neural machine translation. It provides a practical solution for translating Greek text to English, leveraging advanced techniques and datasets.

🚀 Quick Start

Basic Usage

from transformers import FSMTTokenizer, FSMTForConditionalGeneration

mname = "lighteternal/SSE-TUC-mt-el-en-cased"

tokenizer = FSMTTokenizer.from_pretrained(mname)
model = FSMTForConditionalGeneration.from_pretrained(mname)

text = "Ο όρος τεχνητή νοημοσύνη αναφέρεται στον κλάδο της πληροφορικής ο οποίος ασχολείται με τη σχεδίαση και την υλοποίηση υπολογιστικών συστημάτων που μιμούνται στοιχεία της ανθρώπινης συμπεριφοράς ."

encoded = tokenizer.encode(text, return_tensors='pt')

outputs = model.generate(encoded, num_beams=5, num_return_sequences=5, early_stopping=True)
for i, output in enumerate(outputs):
    i += 1
    print(f"{i}: {output.tolist()}")
    
    decoded = tokenizer.decode(output, skip_special_tokens=True)
    print(f"{i}: {decoded}")

✨ Features

Supports Greek to English translation.
Trained using the Fairseq framework with transformer_iwslt_de_en architecture.
Utilizes BPE segmentation (20k codes) and mixed - case model.

📦 Installation

The code example above assumes you have the transformers library installed. You can install it using the following command:

pip install transformers

📚 Documentation

Model Information

Property	Details
Source Languages	el
Target Languages	en
License	apache - 2.0
Dataset	Opus, CCmatrix
Model	transformer(fairseq)
Pre - processing	tokenization + BPE segmentation
Metrics	bleu, chrf

Model description

Trained using the Fairseq framework, transformer_iwslt_de_en architecture. BPE segmentation (20k codes). Mixed - case model.

Training data

Consolidated corpus from Opus and CC - Matrix (~6.6GB in total)

Eval results

Results on Tatoeba testset (EL - EN):

BLEU	chrF
79.3	0.795

Results on XNLI parallel (EL - EN):

BLEU	chrF
66.2	0.623

BibTeX entry and citation info

Dimitris Papadopoulos, et al. "PENELOPIE: Enabling Open Information Extraction for the Greek Language through Machine Translation." (2021). Accepted at EACL 2021 SRW

Acknowledgement

The research work was supported by the Hellenic Foundation for Research and Innovation (HFRI) under the HFRI PhD Fellowship grant (Fellowship Number:50, 2nd call)

📄 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