A Majority Voting System For Reliable Sentiment Analysis Of Product Reviews
Abstract
This article examines the majority voting approach to enhance the reliability of product reviews submitted by users on online platforms and improve the accuracy of sentiment analysis results. The majority voting system determines the final sentiment label by aggregating the results produced by multiple annotators or different models. The advantages of this approach and its application areas (in the annotation process and in combining model outputs) are also examined. The article further details the working principles of the system, the process of evaluating and selecting reviews, and the impact of user votes on reliability. The majority voting approach not only enhances sentiment analysis outcomes but also helps filter out fake or biased reviews. This method is particularly noteworthy as an important methodological tool for improving dataset quality and ensuring trustworthy model outputs in developing ABSA and other linguistics-based models for the Uzbek language.