An Analysis and Comparison Of Deep-Learning Techniques and Hybrid Model for Sentiment Analysis for Movie Review

In today’s world, where each day 2.5 quintillion bytes of data is generated, the safe storage and secure processing of such data is a fundamental task. The vastness of the web has brought trillions of data as comments, reviews, blog posts etc. For a service provider the analysis of these are very essential and also finding out the trends in which sentiment analysis plays a huge role. This differs from typical topic-based text classification as it is done by classifying text based on the sentiment it conveys. User generated text is the basis from which opinion and subject knowledge is drawn for sentiment analysis. In this paper, we deal with the sentiment analysis of the IMDB dataset of movie reviews. We explore the effectiveness of different deep learning techniques such as CNN, LSTM, LSTM-CNN, GRU, BERT, BERT-CNN, BERT-LSTM for the sentiment analysis of the IMDB movie review dataset. Data cleaning and the techniques used are explained and the performance of these various algorithms used are measured in terms of recall, precision and accuracy. In this we hope to find the best machine learning model amongst the ones tested for future research.