TY - GEN

T1 - Using Partial Models to Extract Emotional Estimations from Natural Language Texts

AU - Akhmedov, Ergash Yu

AU - Palchunov, Dmitriy E.

N1 - The study was carried out within the framework of the state contract of the Sobolev Institute of Mathematics (project no. FWNF-2022-0011).

PY - 2024/11/26

Y1 - 2024/11/26

N2 - This article is devoted to the urgent problem of determining the emotional coloring of natural language texts. The article notes that the existing methods for analyzing emotions in text have a number of limitations. The purpose of the article is to develop methods for determining the emotional coloring of texts using partial models. To achieve this goal, the following tasks are solved: creating a specialized data set, training neural networks based on partial models, as well as processing and expanding partial models, building chains of partial models corresponding to sequences of situations for more accurate determination of emotions. The article considers the theory of partial models and its application in the problems of emotional analysis of texts. The concept of an evaluative partial model that formally represents an emotionally colored situation is considered. Particular attention is paid to cases when one situation can evoke several different emotions, possibly opposite in their tonality (ambivalence of emotional assessments). The process of creating a data set with labeled emotional assessments is described. The process of training neural networks based on partial models is considered in detail. The process of analyzing various situations and emotions using trained models is demonstrated. Methods for generating new partial models based on the use of trained neural networks are also described. The advantages of the proposed method are highlighted, including high accuracy of emotion analysis and the possibility of emotional analysis of sequences of situations.

AB - This article is devoted to the urgent problem of determining the emotional coloring of natural language texts. The article notes that the existing methods for analyzing emotions in text have a number of limitations. The purpose of the article is to develop methods for determining the emotional coloring of texts using partial models. To achieve this goal, the following tasks are solved: creating a specialized data set, training neural networks based on partial models, as well as processing and expanding partial models, building chains of partial models corresponding to sequences of situations for more accurate determination of emotions. The article considers the theory of partial models and its application in the problems of emotional analysis of texts. The concept of an evaluative partial model that formally represents an emotionally colored situation is considered. Particular attention is paid to cases when one situation can evoke several different emotions, possibly opposite in their tonality (ambivalence of emotional assessments). The process of creating a data set with labeled emotional assessments is described. The process of training neural networks based on partial models is considered in detail. The process of analyzing various situations and emotions using trained models is demonstrated. Methods for generating new partial models based on the use of trained neural networks are also described. The advantages of the proposed method are highlighted, including high accuracy of emotion analysis and the possibility of emotional analysis of sequences of situations.

KW - LSTM

KW - atomic diagram

KW - estimated partial model

KW - natural language processing

KW - partial model

KW - sentiment analysis

KW - situation representation

UR - https://www.scopus.com/record/display.uri?eid=2-s2.0-85212136823&origin=inward&txGid=448530c0267f70bc675e92685dcac2cb

UR - https://www.mendeley.com/catalogue/94f150e1-b2a7-3f5c-9f47-2d4b1cba6526/

U2 - 10.1109/SIBIRCON63777.2024.10758542

DO - 10.1109/SIBIRCON63777.2024.10758542

M3 - Conference contribution

SN - 9798331532024

T3 - 2024 IEEE International Multi-Conference on Engineering, Computer and Information Sciences, SIBIRCON 2024

SP - 282

EP - 287

BT - 2024 IEEE International Multi-Conference on Engineering, Computer and Information Sciences, SIBIRCON 2024

PB - Institute of Electrical and Electronics Engineers Inc.

T2 - 2024 IEEE International Multi-Conference on Engineering, Computer and Information Sciences

Y2 - 30 September 2024 through 2 November 2024

ER -