Emotional Analysis of Russian Texts Using Emojis in Social Networks

Abstract

Sentiment analysis is a key task in natural language processing and has a wide range of real-world applications. Traditional methods classify “plain texts” as “positive” and “negative”. We propose a method that is significantly different from traditional approaches. In addition to “plain text”, we have analyzed ways to express emotions in a message and a variety of emotional indicators, emoticons and emojis. The proposed model with emotional indicators to predict text polarity improves the prediction accuracy of sentiment classes by 6% compared to traditional models. The model uses an original data set marked up according to an extended list of Plutchik emotion classes. Therefore, the model predicts 8 independent sentiment classes: “joy”, “sad”, “distaste”, “fear”, “anger”, “surprise”, “attention” and “trust”. The novelty of the research also lies in using the Russian language data set. The model considers national linguistic, semantic and semiotic features of social environment.