Enhancing Sentiment Analysis in Uzbek Language Texts through Weighted Lexical Features

Abstract

This article presents an original sentiment analysis algorithm developed for analyzing Uzbek language texts in order to correctly determine and identify the sentiment of the text. The method proposed by the authors provides a rating system where lexical characteristic weights are implemented for each word, phrase, idiom or negation. The proposed system involves marking words with sentiment scores in a numerical range, where the minimum value is -3 and the maximum is +3. The authors also tested the efficiency of the algorithm, where the result showed a high accuracy, and the completeness of the identified objects is in the range from $85 \%$ to $95 \%$, depending on the type of object (word, phrase, negation, etc.). Despite the positive test results, the evaluation process also identified areas for improving the algorithm, including expanding the algorithm’s vocabulary for the highest quality processing of word forms. In addition, authors noted about alternative technologies that might be used for enhancing current algorithm.