The training data is a large dataset of movies followed by a sentiment: pos-if positive neg-if negative sentiment is provided by the review.
This model will take the review of the viewer as input and predicts whether the sentiment is positive pr negative using 'Naive Bayes' classification algorithm.