Sentiment Analysis of IMDB Movie Reviews Using Deep Learning Techniques

Movie reviews help users to evaluate and decide if a certain movie is of their particular interest. Nowadays, there is a lot of data about movies like IMDB which is an extensive database containing thousands of movie reviews. However, analyzing each of these reviews can be time consuming and tedious, so machine learning models could be implemented for automation and analysis of these reviews. Sentiment analysis is a process that uses artificial intelligence and machine learning to find a point of view, a keyword, or a feeling in order to highlight the information of interest in the process. In this sense, an opinion can be interpreted as a dimension in the data regarding a particular topic and can be very useful in various fields of application such as data mining, web mining, and social media analytics. This paper aims to use an IMDB database that contains 50,000 reviews, and we intend to apply transformer-based language models like Bidirectional Encoder Representations from Transformers (BERT), RoBERTa, and XLNet for sentiment analysis. Moreover, we implement a TF-IDF and cluster analysis to gain insights about the topics related to both positive and negative reviews (Yasser in IMDB movie ratings sentiment analysis, 2022 [1]; Kumar et al. in Int J Interact Multimed Artif Intell 5(5), 2019 [2]; Chakraborty et al. in Soc Netw Anal Comput Res Methods Tech 7:127–147, 2018 [3]; Gadekallu et al. in Sentiment analysis and knowledge discovery in contemporary business. IGI Global, pp 77–90, 2019 [4]).