/Sentiment_Prediction_from_Movie_Reviews

Primary LanguagePython

Sentiment_Prediction_from_Movie_Reviews

Sentiment is predicted as either positive or negative on the IMDB dataset.

Dataset-IMDB dataset of 50,000 movie reviews. Words are converted into real-valued vector space using Embedding layer of Keras.

2 models are build-

Simple Multilayer Perceptron Model

  • Embedding layer as the input layer,word vector size of 32.
  • Flatten the Embedding Layer.
  • Dense hidden layer of 250 units with a rectifier activation function.
  • Output layer with one neuron and will use a sigmoid activation function.

1D Convolutional Neural Network

  • Embedding layer as the input layer,word vector size of 32.
  • 1D conv layer ,relu activation,filters 32*3*3.
  • Maxpooling layer .
  • Flatten the layer
  • Dense hidden layer of 250 units with a rectifier activation function.
  • Output layer with one neuron and will use a sigmoid activation function.

Network was trained on Amazon EC2 instances using gpu of g2.2x large.