An intelligent system for ranking e-commerce customer reviews to boost engagement

Abstract

This study introduces an innovative framework employing learning algorithms to effectively rank customer reviews on e-commerce platforms. The approach addresses the inherent ambiguity and subjectivity in customer feedback by leveraging an extensive dataset and sophisticated feature engineering techniques. Central to the methodology is the introduction of an original target variable, the adjusted action rate, which, along with advanced training methods, helps mitigate the prevalent position bias. This is crucial for accurately reflecting the nuances of user behavior and the dynamics of review interaction, ensuring that the most relevant feedback is highlighted for prospective buyers. The framework utilizes Learning to Rank methods specifically designed to tackle the unique challenges of review ranking. These methods prioritize user feedback based on its relevance and helpfulness, enhancing the precision of review rankings. By using advanced machine learning techniques, the framework can discern subtle patterns in user interactions and preferences, providing a more personalized and efficient ranking system. The effectiveness of this approach is evaluated using the Normalized Discounted Cumulative Gain metric, which measures the correlation between user reviews and the AAR. This metric ensures that the ranking system not only improves user satisfaction but also drives engagement performance. The incorporation of regression and classification models further strengthens the framework's ability to handle diverse review data. Regression models predict the adjusted action rate by analyzing various features derived from the reviews, such as length, sentiment, and user credibility. Classification models, on the other hand, help categorize reviews based on their relevance, ensuring that the most significant feedback is prioritized. These models collectively enhance the accuracy and reliability of the review ranking system, making it more robust and adaptive to different user needs. Validation demonstrates significant improvements in user interaction and decision-making efficiency, enhancing the shopping experience by enabling customers to access the most relevant reviews. Detailed analysis reveals substantial increases in key engagement metrics, confirming the model's robustness and reliability. The framework successfully addresses the complexities of review ranking, benefiting both users and vendors. Its ability to adapt to evolving user preferences by continuously learning from new data ensures that the most current and relevant reviews are highlighted, keeping the platform dynamic and user-centric. This adaptability enhances user satisfaction and fosters greater trust in the e-commerce platform, providing accurate and helpful feedback consistently.