Multidimensional dynamic monitoring of product attributes integrating textual semantic features

Text-based quality management often relies on static, retrospective analysis of product attributes, limiting its ability to capture dynamic quality changes. Although statistical process monitoring (SPM) enables real-time tracking, its integration with text analytics has mainly focused on service quality, leaving product-attribute monitoring underexplored. To address this gap, this study proposes a multidimensional framework for dynamic product-attribute monitoring. An Attribute-Weighted Word Mover’s Distance (AW-WMD) is first developed to transform complaint texts into a semantic distance measure, which is jointly monitored with complaint intervals. A distribution-free bivariate exponentially weighted moving average scheme based on tri-aspect rank statistics (BEWMA-TRS) is then proposed to detect location, scale, and shape shifts. By monitoring distinct product subsystems independently, the framework enables component-level anomaly localisation. Simulation results and an empirical case demonstrate that the proposed scheme provides competitive and robust overall performance compared with existing benchmarks, offering a rigorous framework for real-time quality diagnosis and continuous improvement.