Nowadays, developing and testing different models for a natural language processing problem is
an interesting and challenging task. However, due to the nature of the problem, the accuracy of
sentiment analysis on single sentence like movie reviews never reaches above 80% for the past
7 years [1]. Looking at last years project on twitter [2], their accuracy was 59.32% to 63.71%,
depending on different models. In our project, we achieved near 20% more than their result, which
is a significant improvement.