A survey of sentiment analysis from film critics based on machine learning, lexicon and hybridization

Recent research and developments in the field of Sentiment Analysis (SA) have made it possible to simplify the detection and classification of sentiments from the textual content. This type of analysis classifies the text according to its positive, negative, or neutral polarity. Recently, researchers have focused on film reviews and aim to extract personal information about text reviews that can, for example, be used to determine the listener’s position on a number of different topics. The main contributions, proposed in the literature, focused on three categories of approaches: (i) a first category based on the lexicon, (ii) a second category based on machine learning, and (iii) a third based on a hybridization of the two previous categories. To our knowledge, and until the elaboration of this study, no previous study has examined the approaches and levels of sentiment in the field of film reviews. In this article, we propose to review and analyze the main works in this field. We begin by giving a methodological review of our study. Then, we present a taxonomy on the domain of sentiment analysis and a generic view of the main families of sentiment classification techniques. As a next step, we describe the different levels of sentiment analysis considered in the literature, then we expose the process of pre-processing, extracting, and selecting the characteristics necessary for the sentiment analysis. We then propose an analysis and a discussion of the results of the main works studied on sentiment analysis. This presentation will then be followed by a discussion of some research questions and a proposal for a number of future directions in this area that we believe are essential to contribute to solving the problem addressed in this article.