Algorithm of decision trees ensemble for sentiment analysis of Uzbek text

The rapid development of social networks is leading to the growth of the content increased there. Determining the emotional trend of commentaries in this content through natural language processing and sentiment analysis technologies is so beneficial for the timely understanding of community views on social media, tracking brands, and consumer support. In this paper, an algorithm for constructing an ensemble of decision trees for the sentiment analysis of texts written in the Uzbek language was proposed. The algorithm is based on constructing mutually independent trees and making the final decision based on their predictions. The developed algorithm was experimentally researched on the corpus of Uzbek movie reviews. The comparative analysis and effectiveness of the proposed algorithm compared to other machine learning algorithms in the sentiment analysis of Uzbek texts were demonstrated.

Перевод пока недоступен