Machine-Translation

Machine Translation using LSTM Sequence to Sequence Model in Keras

Model Architetcure:

1. Encoder :

• Encoder will take in the input as english sentence and return a hidden state

2. Decoder :

• Decoder takes the input as hidden state from encoder and try to predict marathi transalation of english sentence

Dataset :

Click here to download the dataset. It consist of total 35832 sentences which are translated from english to marathi

Train Model:

python train.py

Note: If code throws a MemoryError, then one hot encode the output in batches

Inference Model:

python Inference_Model.py

Note: I have saved the data in pickle file and load it. you need to first save the data in pickle file,so uncomment the line #13 and #14

Accuarcy after 30 epochs: