Sentiment Quantification of User-Generated Content

User-generated content (UGC) has turned into a goldmine for market researchers, social scientists, political scientists, and reputation managers, since it gives near-instant access to a potentially enormous quantity of data from which the collective sentiment about products, companies, policies, and political candidates, can be gauged. Possibly the most important task underlying efforts to tap into this goldmine is sentiment classification, the task of classifying an item of UGC (e.g., a tweet, a product review, a post on a social networking service) according to the sentiment it conveys (or opinion it expresses) about a certain entity or topic. However, it turns out that, in many applications, the final goal of sentiment classification is not that of determining the class of individual UGC items, but that of computing the percentage of UGC items that belong to a certain class. When the latter task is tackled via supervised learning, it is known as sentiment quantification. We here give an introduction to the task of quantifying UGC by sentiment, to the methods that have been proposed in the literature, and to the measures that are used for evaluating the accuracy of these methods.