Overview / Usage

Training a long short term memory recurrent network on a dataset of labeled movie reviews to learn how to predict whether some text is either positive or negative.

Methodology / Approach

1. Collect text data (Movie reviews).
2. Convert input review text into encoded vectors using an embedding layer.
3. Input these word embeddings into the LSTM cells (allow for sequence learning).
4. Pass LSTM outputs to a sigmoid output layer (positive or negative).
5. Compute loss by comparing LSTM output to actual label, use calculus to optimize the network during training.

Technologies Used

Pytorch

Repository

https://github.com/Amir22010/NLP_Deep_Learning/tree/master/7_June_2019_NLP_Deep_Sentiment

Other links

• Project Code (https://github.com/Amir22010/NLP_Deep_Learning/tree/master/7_June_2019_NLP_Deep_Sentiment)