Dual-Track Lifelong Machine Learning-Based Fine-Grained Product Quality Analysis

Artificial intelligence (AI) systems are becoming wiser, even surpassing human performances in some fields, such as image classification, chess, and Go. However, most high-performance AI systems, deep learning models, black boxes (i.e., only system inputs outputs visible, but the internal mechanisms unknown) and, thus, notably challenging to understand. Thereby a with better explainability is needed help humans understand AI. This paper proposes dual-track approach that uses reinforcement supplement fine-grained learning-based sentiment classification. Through lifelong machine learning, can gradually become wiser realize high performance (while keeping outstanding explainability). The extensive experimental results show proposed provide reasonable analyses product reviews remarkably achieve 133% promotion of Macro-F1 score on Twitter classification task 27.12% an Amazon iPhone 11 task, respectively.