A Vietnamese-English Neural Machine Translation System

Our pre-trained VinAI Translate models vinai/vinai-translate-vi2en and vinai/vinai-translate-en2vi are state-of-the-art text-to-text translation models for Vietnamese-to-English and English-to-Vietnamese, respectively. These pre-trained models are currently used in the translation component of our VinAI Translate system. The general architecture and experimental results of the pre-trained models can be found in our VinAI Translate system paper:

@inproceedings{vinaitranslate,
title     = {{A Vietnamese-English Neural Machine Translation System}},
author    = {Thien Hai Nguyen and 
             Tuan-Duy H. Nguyen and 
             Duy Phung and 
             Duy Tran-Cong Nguyen and 
             Hieu Minh Tran and 
             Manh Luong and 
             Tin Duy Vo and 
             Hung Hai Bui and 
             Dinh Phung and 
             Dat Quoc Nguyen},
booktitle = {Proceedings of the 23rd Annual Conference of the International Speech Communication Association: Show and Tell (INTERSPEECH)},
year      = {2022}
}

Please CITE our paper whenever the pre-trained models or the system are used to help produce published results or incorporated into other software.

Using VinAI Translate in transformers

Installation

pip install transformers sentencepiece tokenizers

Pre-trained models

Model	#params	Max length
vinai/vinai-translate-vi2en	448M	1024
vinai/vinai-translate-en2vi	448M	1024

• Users might also play with these models via the VinAI Translate text-to-text translation app at: https://huggingface.co/spaces/vinai/VinAI_Translate

Vietnamese-to-English translation

from transformers import AutoTokenizer, AutoModelForSeq2SeqLM

tokenizer_vi2en = AutoTokenizer.from_pretrained("vinai/vinai-translate-vi2en", src_lang="vi_VN")
model_vi2en = AutoModelForSeq2SeqLM.from_pretrained("vinai/vinai-translate-vi2en")

def translate_vi2en(vi_text: str) -> str:
    input_ids = tokenizer_vi2en(vi_text, return_tensors="pt").input_ids
    output_ids = model_vi2en.generate(
        input_ids,
        do_sample=True,
        top_k=100,
        top_p=0.8,
        decoder_start_token_id=tokenizer_vi2en.lang_code_to_id["en_XX"],
        num_return_sequences=1,
    )
    en_text = tokenizer_vi2en.batch_decode(output_ids, skip_special_tokens=True)
    en_text = " ".join(en_text)
    return en_text

vi_text = "Cô cho biết: trước giờ tôi không đến phòng tập công cộng, mà tập cùng giáo viên Yoga riêng hoặc tự tập ở nhà. Khi tập thể dục trong không gian riêng tư, tôi thoải mái dễ chịu hơn."
print(translate_vi2en(vi_text))

vi_text = "cô cho biết trước giờ tôi không đến phòng tập công cộng mà tập cùng giáo viên yoga riêng hoặc tự tập ở nhà khi tập thể dục trong không gian riêng tư tôi thoải mái dễ chịu hơn"
print(translate_vi2en(vi_text))

• NOTE: Before training, we performed Vietnamese tone normalization on the Vietnamese training data, using a Python script available HERE. Users should also employ this tone normalization script to pre-process the Vietnamese input data before feeding the data into our pre-trained model vinai/vinai-translate-vi2en.

English-to-Vietnamese translation

from transformers import AutoTokenizer, AutoModelForSeq2SeqLM

tokenizer_en2vi = AutoTokenizer.from_pretrained("vinai/vinai-translate-en2vi", src_lang="en_XX")
model_en2vi = AutoModelForSeq2SeqLM.from_pretrained("vinai/vinai-translate-en2vi")

def translate_en2vi(en_text: str) -> str:
    input_ids = tokenizer_en2vi(en_text, return_tensors="pt").input_ids
    output_ids = model_en2vi.generate(
        input_ids,
        do_sample=True,
        top_k=100,
        top_p=0.8,
        decoder_start_token_id=tokenizer_en2vi.lang_code_to_id["vi_VN"],
        num_return_sequences=1,
    )
    vi_text = tokenizer_en2vi.batch_decode(output_ids, skip_special_tokens=True)
    vi_text = " ".join(vi_text)
    return vi_text

en_text = "I haven't been to a public gym before. When I exercise in a private space, I feel more comfortable."
print(translate_en2vi(en_text))

en_text = "i haven't been to a public gym before when i exercise in a private space i feel more comfortable"
print(translate_en2vi(en_text))