Skip to content
Publications

Talk · Ural Symposium on Biomedical Engineering, Radioelectronics and Information Technology · May 14, 2026

Stages of Developing Automated Lexical Resources for Uzbek Language Sentiment Analysis

Samida T. Mustafaeva, Gulshoda Abdumalikova, Dilnoza Jamolitdinova

Abstract

This paper aims to develop a methodology for the autonomous generation and expansion of a sentiment lexicon for the Uzbek language, addressing the limitations of manual dictionary compilation in a low-resource, agglutinative linguistic environment. The proposed model integrates web scraping, AI-powered labeling, and linguistic preprocessing (lemmatization). The core framework combines distributive semantics (Word2Vec/FastText) with Pointwise Mutual Information (PMI) for statistical sentiment weighting. A multi-genre corpus of $\mathbf{1 0, 0 0 0}$ sentences was used for empirical validation. The automated approach proved $\mathbf{1, 0 0 0}$ times more efficient than manual compilation, achieving over 85% accuracy. The algorithm successfully identified 3,000-5,000 sentiment-bearing lemmas, though challenges remains regarding irony and polysemy. The research introduces a ‘dynamic lexicon’ framework for low-resource agglutinative languages, providing a scalable foundation for social monitoring, business intelligence, and public administration systems in Uzbekistan.

Sentiment Analysis and Opinion MiningNatural Language Processing TechniquesHate Speech and Cyberbullying Detectionsentiment lexiconUzbek languagesentiment analysis

More by Dilnoza Jamolitdinova