Building the Plain RNN Model

In the next exercise, we will build our first model for the sentiment classification task using plain RNNs. The model architecture we'll use is depicted in the following figure, which demonstrates how the model would process an input sentence "Life is good", with the term "Life" coming in at time step T=0 and "good" appearing at time step T=2. The model will process the inputs one by one, using the embedding layer to look up the word embeddings that will be passed to the hidden layers. The classification will be done when the final term, "good", is processed at time step T=2. We'll use Keras to build and train our models:

Figure 6.4: Architecture using an embedding layer and RNN

Exercise 6.01: Building and Training an RNN Model for Sentiment Classification

In this exercise, we will build and train an RNN model for sentiment classification. Initially, we will define the architecture...