English - Indic -English Transliteration

This project is focused on word to to word transliteration from English to Indic languages and vice-versa. . This was done using seq2seq architecture using LSTM and GRU and LSTM with Bahdanau Attention mechanism.

Requirements

• requirements.txt

Table Of Contents

• In a Nutshell
• In Details
• Future Work
• Contributing
• Acknowledgments

In a Nutshell

1. We gathered the dataset(text) from Internet. You can go the input folder to get the dataset.
2. We have written code for seq2seq models using LSTM, GRU and LSTM with attention which can be found in src folder.
3. We then trained 8 models(4 for Eng-Indic & 4 for Indic-Eng) for each architecture here. The models can be found in models folder.
4. We have then hosted the webapp in streamlit. Link for the website to try it out.

In Details

├──  input
│    └── *.xml - datasets for English to Indic Languages.
│
│
├──  models  
│    └── *  - 8 models for 3 different architecture(LSTM, GRU and LSTM_attn).
│ 
│
├──  src
│    └── config.py  - configuration for different languages
│    └── dataset.py - dataset generator
│    └── language_preprocessing.py - preprocessing text
│    └── train.py - to train different models
│    └── webapp.py- streamlit webapp
│    └── gru.py- gru model
│    └── lstm.py- lstm model
│    └── lstm_attention.py- lstm_attn model

Future Work

This can be extended to transformer based models as well. Currently working on it.

Contributing

Any kind of enhancement or contribution is welcomed.