In this section, we will use LSTM networks to perform sentiment analysis. Along with the word itself, the LSTM network also accounts for the sequence using recurrent connections, which makes it more accurate than a traditional feed-forward neural network.

Here, we shall use the movie reviews dataset text2vec from the cran package. This dataset consists of 5,000 IMDb movie reviews, where each review is tagged with a binary sentiment flag (positive or negative).