opennmt-en-hu

Introduction

English - Hungarian translation model that was trained on the Hunglish2 dataset using OpenNMT.

Usage

Install the necessary dependencies:

pip3 install ctranslate2 pyonmttok

Simple tokenization & translation using Python:

import ctranslate2
import pyonmttok
from huggingface_hub import snapshot_download
model_dir = snapshot_download(repo_id="SZTAKI-HLT/opennmt-en-hu", revision="main")

tokenizer=pyonmttok.Tokenizer(mode="none", sp_model_path = model_dir + "/sp_m.model")
tokenized=tokenizer.tokenize("Hello világ")

translator = ctranslate2.Translator(model_dir)
translated = translator.translate_batch([tokenized[0]])
print(tokenizer.detokenize(translated[0].hypotheses[0]))

Citation

If you use our model, please cite the following paper:

@inproceedings{nagy2022syntax,
  title={Syntax-based data augmentation for Hungarian-English machine translation},
  author={Nagy, Attila and Nanys, Patrick and Konr{\'a}d, Bal{\'a}zs Frey and Bial, Bence and {\'A}cs, Judit},
  booktitle = {XVIII. Magyar Számítógépes Nyelvészeti Konferencia (MSZNY 2022)},
  year={2022},
  publisher = {Szegedi Tudományegyetem, Informatikai Intézet},
}