| Abstract |
Nowadays, several platforms on the web and social networks like Facebook, Twitter, IMDB (Internet Movie Database) propose to share feelings and opinions on a variety of topics. This information is very important in several fields like policy, digital marketing, social or individual, and their analysis allow as to extract opinions and to determine the subjective information contained in the texts. However, the sentiment analysis area is confronted to several problems that distinguish it from traditional thematic research, since the sentiment is expressed in a very varied and very subtle ways. In the last few years, several researches focused on sentiment analysis in order to study attitudes, opinions, and emotions. In this paper we present an analytical and comparative study of different researches conducted on sentiment analysis in social networks. Our comparison takes into account variants criteria such as: the research objective, the language nature, the preprocessing steps, the approach used for sentiment classification…. This study analyzes in more detail the preprocessing steps which are very important in sentiment analysis process success and are the most difficult especially in the case where the comments are written in not structured language. © 2019 Association for Computing Machinery. |
