S2SLT-E2K: Sequence to Sequence Language Translation from English to Konkani using Natural Language Processing

• Tech Stack: Python, Tensorflow, Keras, Pandas, Microsoft Azure

We have taken dataset from TDIL website. Quotes were taken from Shabdkosh website. The dataset consists of 70000 English-Konkani sentence. Using libraries like Tensorflow, Keras, Pandas, Sklearn we have done preprocessing of dataset and then we used Sequence-to-Sequence (Encoder-Decorder) with the help of Long Short Term Memory (LSTM) for mapping English word with Konkani words and finally we combined the sentence for getting final output. We have also done tuning of hyper-parameters, optimiser, and regularisation techniques. We got 30.05 BLEU Score and an accuracy of 63.62%.