Since the advent of Web 2.0, people have become more interested in expressing and sharing their views about everyday life and global issues on the Web. The evolution of social networks has also contributed greatly to these activities, thereby providing a worldwide transparent opinion-sharing platform. Such electronically expressed comments are very common in business and service industries giving customers a chance to voice their opinions. Research communities, academics, and general and service industries have seriously been working on sentiment analysis over the past fifteen years to extract and analyze public views and attitudes for the purpose of opinion mining. Sentiment analysis is a research area emerging as a result of rapid developments in computer science. The growing amounts of user-generated content on the Web have turned sentiment analysis into an important tool for the extraction of information regarding the human sentiment status. This paper conducts a review analysis to identify developments and shortcomings in sentiment analysis as well as the effective factors. For this purpose, a comparative review of different sentiment analysis methods is carried out. The relevant developments are described by classifying studies into four dimensions, i.e. machine learning methods, statistical techniques, evolutionary algorithms, and use of language datasets. This paper presents an accurate review of sentiment analysis depicting the views of more than forty papers on the necessary tasks, approaches, and applications of sentiment analysis over the past decade. According to the reviews of deep learning approaches to machine learning, an outlook is provided to properly achieve goals by using the functionality for the extraction of syntactic and semantic features from a text.
Azgomi et al. (Fri,) studied this question.