Computational Approaches to Sentiment Estimation in Uzbek Essays

Abstract

This paper examines the problem of sentiment detection in Uzbek essays. The goal is simple: to classify the text into one of three classes-negative, neutral, or positive. All experiments use minimal preprocessing, including only alphabetical standardization and apostrophe unification. Three statistical approaches are compared: a method based on n-gram expansion without model training, a method based on TF-IDF n-grams with a linear model, and a method based on a support vector machine. Quality is assessed using Accuracy, Macro-F1, and a confusion matrix. Robustness testing is also conducted on three thematic datasets of 100 essays each. The results show that trained methods based on TF-IDF features perform more reliably across topics, while SVM produces the most consistent quality and copes better with ambiguous texts. Neutral remains the most challenging class, as such texts often contain mixed evaluations. Meanwhile, in addition to the main technical component, the authors of the article specifically focus on the linguistic aspects of the Uzbek language. Specifically, they examine in detail such aspects as the nature of the language, its morphology, and syntax, with the aim of taking these properties into account when developing the corresponding algorithms.

More by Nilufar Abdurakhmonova

• A model of a universal linguistic knowledge graph for Turkic languagesArctic XXI century · 2026
• Machine Learning-Based Topic Classification of Economic and Financial Texts in UzbekInternational Conference on Computer Science and Engineering · 2026
• Building a Phraseological Ontology of Uzbek Idioms and Evaluating Similarity Measures for Their Detection in TextInternational Conference on Computer Science and Engineering · 2026
• Developing Morphological and Syntactic Analysis Models for Karakalpak as a Low-Resource Turkic LanguageInternational Conference on Computer Science and Engineering · 2026