The objective of this project was to implement sentiment analysis on human written movie reviews, using deep-learning models. To determine whether the given movie review has a positive or negative sentiment, two different models were developed, one BiLSTM and one CNN.
At the beginning, simple versions of them were established and used as baseline models. Subsequently, a number of modifications were tried and added on top of these baselines, in order to achieve better accuracy on the classification of every review as positive or negative.
In ths report, the detailed procedure of development and experimental phases will be presented, followed by the final conclusions and results.