Abstract
The paper presents one of the first stages of the project which aims to design a classifier of Internet texts in Russian by emotional tonality. The relevance of the study not only consists in the Russian language data, but also in the attempt of creating a multiclass classifier. For the most accurate realization of the idea of the project, the Lövheim cube model including eight emotions was chosen, as well as methods that tend to increase the objectivity of the results, such as an independent assessors’ data labeling, a use of corpus linguistics tools, etc. In previous linguistic analysis of the dataset taken from social network VKontakte and subsequent validation of features of emotions with the help of the corpus manager and the prototype of the classifier it has been stated that the classes Shame and Disgust, unlike other classes, did not demonstrate any peculiarities regarding lexico-morphological level. Due to the hypothesis of the reliability of syntactic features of these classes, as well as the fact of them being characterized as low-noradrenaline emotions, it has been suggested that these two must be correlated. Methods of contextual analysis, corpus linguistics and statistics have proved the relevance of specific syntactic configurations as predictors of Shame and Disgust in the Internet texts in Russian, for instance, “subject in dative case with a verb было and adverb” or “subject with a verb быть in an appropriate form and an adjective”. The syntactic features of Shame and Disgust could be used as emotional classifier parameters.
Highlights
The mentioned above signal substances or neurotransmitters form the axes of a coordinate system, and eight basic emotions are placed in the eight corners of cube model
To deal with emotional text classes of Shame and Disgust we focused on the syntactic specificity of texts colored by emotions of Shame and Disgust
The applied algorithm including the steps on different levels of linguistic abstraction and largely supported by corpus tools has allowed us to detect some syntactic features of “ashamed” and “disgusted” Internet texts in Russian
Summary
The article presents some results of the project conducted on the field of sentiment analysis and supposed to resolve the problem of attributing Internet text in Russian to the particular class of emotions. the technologies of binary and ternary sentiment analysis are rather developed (Nakov, Ritter, Rosenthal, Sebastiani, & Stoyanov, 2016; Yang, He, & Chen, 2019; Yousefpour, Ibrahim, & Hamed, 2017), the multiclass emotional text classifier represents yet a new task worth to be accomplished.Our aim is to run a computer classifier able to detect and define the emotion mostly represented in the Internet text in Russian or to attribute the text to the emotionally neutral class.For this purpose, we use the classification of emotions proposed by Swedish neuroscientist H. The article presents some results of the project conducted on the field of sentiment analysis and supposed to resolve the problem of attributing Internet text in Russian to the particular class of emotions. Our aim is to run a computer classifier able to detect and define the emotion mostly represented in the Internet text in Russian or to attribute the text to the emotionally neutral class. For this purpose, we use the classification of emotions proposed by Swedish neuroscientist H. The mentioned above signal substances or neurotransmitters form the axes of a coordinate system, and eight basic emotions are placed in the eight corners of cube model. Lövheim uses the first for denoting the softer expression of the emotion, the second – for the stronger: Interest / Excitement, Enjoyment / Joy, Surprise, Distress / Anguish, Anger / Rage, Fear / Terror, Contempt / Disgust, Shame / Humiliation
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
