Method of Recognizing Attitude Components in Online Discussions Based on Neural Network Models

Авторы

Девяткин Д. А. ,  Кузнецова Ю. М. ,  Чудова Н. В. ,  Чуганская А. А. ,  Киреев Д. А.

Аннотация

Modern research on large social groups has focused on the use of big data. One of the significant areas of study is the analysis of attitudes and their components. The goal of this work was to explore the possibility of creating an automated tool for recognizing attitude components in online discussions. The hypothesis was that there would be a difference in the effectiveness of recognizing attitude components using neural network models pre-trained on a Russian-language corpus of online discussions. The study was based on psycholinguistic methods and language model training in the study of a large social group of users of Internet discussions. Neural network models of the BERT and ELECTRA families were used. The accuracy and F1 are used to compare models, and it was maximized, when selecting the hyperparameters. The ruRoBERTa-large model achieved the best results with an F1 of 75.97% and an accuracy of 72.69%. Applying machine learning methods based on pre-trained transformer language models to a corpus of texts in which social media users express their attitudes towards social objects allows us to simulate the work of a psychologist who evaluates the representation of attitude components in the statements with an acceptable level of accuracy.

Внешние ссылки

DOI: 10.1007/978-3-032-39382-1_6

Скачать статью (PDF) на сайте конференции (англ.): https://icpr2026orgteam.github.io/WSpapers/pdfs/paper_0053.pdf

ResearchGate: https://www.researchgate.net/publication/415064333_Method_of_Recognizing_Attitude_Components_in_Online_Discussions_Based_on_Neural_Network_Models

Ссылка при цитировании

Chuganskaya, A., Kireev, D., Chudova, N., Devyatkin, D., Kuznetsova, Y. (2027). Method of Recognizing Attitude Components in Online Discussions Based on Neural Network Models // In: Sintorn, IM., Pal, U., Lu, S. (eds) Pattern Recognition. ICPR 2026 International Workshops. ICPR 2026. Lecture Notes in Computer Science, vol 17111, pp. 85–93. https://doi.org/10.1007/978-3-032-39382-1_6