AmazonReview-Sentiment-Analysis

Sentiment Analysis using simple LSTM and Conv1D layers.

This is model that is used to get the sentiment rating(0-1 range). 0 means bad/worse and 1 means good sentiment.

Dependencies

Python 3+
Keras with tensorflow backend
nvidia gpu (for training purpose as it use CuDNNLSTM layer that is accelerated by CuDNN library by nvidia)
Numpy

How to use

Fork this repo
Download the dataset from here .
Download the GloVe Word embeddings from here.
Save both data and GloVe embeddings in data folder.
If training, make changes in file utils/config.py if you want. Change TRAINING_DATASET_SIZE accordingly to change the number of training sentences.
Use the train.ipynb notebook for training.
If using for test-predictions, download the weights from here and save it in weights folder.
Use inference.ipynb notebook.