This project focuses on Solving Sequence-to-Sequence problem of Machine translation which is conversion of English words into corresponding Hindi words by the means of a Encoder-Decoder Architecture.

The work flow is as follows -

• Preprocessing - The inputs are converted into suitable format along with a start and end token so that it can be fed to the encoder.
• Encoder - This is the initial part of the Encoder decoder structure where the networks learn to predict the initial texts and weights are updated.
• Decoder - This part translates the encoded weights into the target language by utilizing the weights provided by the encoder structure.

The encoder-decoder model is a way of using recurrent neural networks for sequence-to-sequence prediction problems. The approach involves two recurrent neural networks, one to encode the input sequence, called the encoder, and a second to decode the encoded input sequence into the target sequence called the decoder. The pretrained model is available in the code section

The Dataset used is englishHindicorpora- https://www.kaggle.com/aiswaryaramachandran/hindienglish-corpora Which Consists of Pairs of hindi and English sentences.

These are some of the results of the notebook -

• Numpy - To perform Scientific mathetatics
• Tensorflow 2.x - A deep learning library
• Pandas - For manipulating data

• Improve model performance by using more training examples
• Implement Attention mechanism for better performance
• Using Pretrained vectors for better results

want to contribute ? contact me at [email protected]