Neural Translate Machine is an open-source project aimed at providing English-to-Vietnamese translation using state-of-the-art neural network models. It leverages the T5 model as the main contributor, along with three Transformer-based models and ConvS2S as a baseline.
- English-to-Vietnamese translation
- Multiple neural network models, including T5, Transformer, and ConvS2S
- Training on the iwlst dataset
- Open-source project for contributions and further development
The T5 model, short for "Text-to-Text Transfer Transformer," is a versatile and powerful neural network architecture for natural language processing tasks. It employs a transformer-based architecture with 12 encoder-decoder layers, attention mechanisms, and positional encodings. T5 is pre-trained on a massive amount of text data and can be fine-tuned for specific downstream tasks like translation.
To run the T5 model, please refer to the Transformer.ipynb notebook available on Kaggle. For the Transformer-FairSeq and ConvS2S models, please use Google Colab.
The T5 model achieves a Multi bleu score of 31.5 for English-to-Vietnamese translation.
No specific dependencies or prerequisites are required for running the provided models.
Contributions to this project are welcome. You can contribute by continuing training on other datasets, improving the existing models, or adding new features to enhance translation accuracy and performance.
For any inquiries or feedback, please reach out to us via email at trinhtan339@gmail.com.
This project is released under an open-source license.