Skip to content

Latest commit

 

History

History
32 lines (23 loc) · 828 Bytes

File metadata and controls

32 lines (23 loc) · 828 Bytes

Low-Resources-Machine-Translation

Implementing a Machine translation system from English to French for the course IFT-6759

  1. To train a en-fr model.
python train.py --epoch=10 --seed=1234 
  1. To train a reverse(fr-en) model and generate files for back-translation
python train.py --epoch=10 reverse_translate=True generate_samples=True --seed=1234  
  1. To use these additional files created above.
python train.py --epoch=10 add_synthetic_data=True --seed=1234  
  1. To create your own word_embeddings.
python word2vec.py --lang=en --emb-dim=256 --vocab-size=20000  

Configure config.py for more control over all parameters and paths.

Steps for the evaluation:

python evaluator.py --input-file-path="path_to_en_file.txt" --target-file-path="path_to_org_fr_file.txt"