An enhanced approach for movie review analysis using deep learning techniques

The principle thought behind the sentiment analysis is to predict the emotions or opinions of the general population towards a specific subject from organized, semi-organized or unstructured literary information. With the rapid growth of internet, people around the world are increasingly passionate about expressing their emotions and opinions on the internet. Most of the industries and organizations formulate significant decisions based on customer reviews available on social websites. Due to the fast advancement in text data processing, sentiment analysis has turned into a significant research direction in Natural language processing (NLP). In recent days, the popularity of the deep learning techniques is also gaining attention which in turn contributes to enhancements in existing sentiment analysis techniques. In this paper, we perform analysis of movie reviews using a combination of deep learning techniques, LSTM and CNN. CNN is a commonly used deep learning method used for feature learning. LSTM is a unique sort of RNN, which is equipped for adapting long term conditions. This model allows remembering of information over a much longer period. We apply two different deep learning models to classify IMDB movie reviews. The first model is a hybrid LSTM- CNN and second is hybrid CNN-LSTM. We found that LSTM-CNN model outperformed CNN-LSTM model with an overall accuracy of 79%.