Identifying Product Aspect Polarity by Product Review Classification with Dual Sentiment Analysis

Abstract

Dual Sentiment Analysis has emerged as a crucial and active research field. It involves extracting sentiment from comments, feedback, or critiques, which serves as valuable indicators for various purposes. To address this, we propose a novel dual training algorithm that utilizes both original and reversed training reviews to develop a robust sentiment classifier. Additionally, we introduce a dual prediction algorithm that comprehensively assesses both aspects of a review for classification during testing. The proposed approach goes beyond traditional polarity (positive-negative) classification by extending the framework to a 3-class system, which includes neutral reviews. This enhancement allows for a more nuanced understanding of sentiment. By considering neutral reviews, we gain deeper insights into the sentiment landscape. Dual Sentiment Analysis plays a pivotal role in helping companies gauge the level of acceptance of their products and formulate strategies to improve product quality. Moreover, it empowers policymakers and politicians to gain