In this paper, we attempted to find speech features of different reactions to frustration to detect and classify them in social media texts. Frustration is a highly motivated situation in which it is impossible to achieve a goal when unexpected external or internal obstacles are encountered to meet the need. We use a well-recognized typology of the reactions and focus on context-aware but straightforward models and classification features, which can be easily interpreted. The experiments show that pure lexis cannot be used as the only feature for the classification. Only the models, which combine different-level linguistic features, implicitly like in BERT or in the models with the linguistic patterns, provide fair results. From a psychological point of view, some misclassifications of the obtained reaction data can be related to their assignment to one class of extrapunitive reactions. Discussions in social networks suggest a high level of human activity, a desire to seek a solution to the problem in a broader social interaction. Thus, the focus on extrapunitive reactions and an increased emotional component in the form of aggression is a feature of that interaction type. On the one hand, we provide a method to classify the social network messages; on the other hand, the training results can be interpreted and analyzed by experts in psychodiagnostics.
Read at Google Books: https://books.google.ru/books?hl=en&lr=&id=_SJGEAAAQBAJ&oi=fnd&pg=PA17
Microsoft Academic: https://academic.microsoft.com/paper/3203578942/
Semantic Scholar: https://api.semanticscholar.org/CorpusID:238416709
Devyatkin D., Chudova N., Chuganskaya A., Sharypina D. (2021) Methods for Recognition of Frustration-Derived Reactions on Social Media. In: Kovalev S.M., Kuznetsov S.O., Panov A.I. (eds) Artificial Intelligence. RCAI 2021. Lecture Notes in Computer Science, vol 12948. Springer, Cham. Pages 17-30