Sentiment Analysis. Current Approaches and New Directions

Today, the Web has become an essential source of information, thanks to the quantity and diversity of textual content expressing the opinions of Internet users. This content is manifold: blogs, comments, forums, social networks, reactions or opinions, increasingly centralized by search engines. Given this abundance of data and sources, the development of tools to extract, synthesize and compare the opinions expressed on a given subject is becoming crucial. This type of tool is of considerable interest to companies seeking customer feedback on their products or brand image, as well as to individuals seeking information for a purchase, an outing or a trip.

It was in this context that opinion analysis (commonly known as sentiment analysis or opinion mining) was born. The first studies in automatic opinion mining date back to the late 1990s, with a particular focus on determining the polarity of adjectives in documents, i.e. whether the opinion conveyed by the adjectives is positive or negative. Since the 2000s, a large number of works have been published on the subject, making opinion extraction one of the most active fields in Automatic Language Processing (ALP) [H7258] and data mining, with over 26,000 publications listed on Google Scholar. It's important to note that, before it became a field of computer science research, opinion analysis was widely studied in linguistics , psychology , sociology and economics