Integration of a Sentiment Dictionary and Named Entity Recognition System into a Tool for Analyzing Public Opinion in the Uzbek Language
Abstract
This article describes the development and implementation of a tool for analyzing public opinion in the Uzbek language. The approach is based on combining methods for analyzing text tone with named entity recognition technologies. The system architecture includes several sequential stages: preliminary text processing, formation of an emotional coloring dictionary, training models for identifying named objects, and integration of all modules into a single analytical platform. The developed tool makes it possible not only to identify the general emotional background of a text, but also to link the obtained assessments to specific objects— individuals, organizations, products, and geographical names. The practical value of the solution is confirmed by examples of monitoring news sources and analyzing user reviews. The results show that combining lexical methods with machine learning algorithms provides a deeper understanding of context and improves the accuracy of tone detection in Uzbek-language texts. This approach may be in demand by government agencies, commercial structures, and sociological services for analyzing the media space and feedback.