Used natural language processing techniques to convert around 50,000 review texts to a form suitable for LSTM to recognise like tokenization and padding and then trained a bidirectional LSTM Recurrent Neural Network on around 38,000 movie review texts to recognise whether a given movie review is positive or negative and output the corresponding sentiment score between 0 and 1. Highly positive reviews were given sentiment score close to 1 and highly negative reviews were given score close to 0. Moderate reviews were given score close to 0.5. Then, tested the model for new unknown reviews downloaded from the internet and got around 90% accuracy.
kshitijshrivastava1903/Sentiment-Classification-Movie-Review-LSTM-NLP
Trained a bidirectional LSTM Recurrent Neural Network on around 38,000 movie review texts to recognise whether a given movie review is positive or negative and output the corresponding sentiment score between 0 and 1.
Jupyter Notebook