While consumer complaints are recognized as the primary catalyst for product recalls in numerous sectors with high recall rates (such as automobiles, food, beverages, and phar-maceuticals), both firms and regulatory bodies face challenges due to limited human and technological resources when it comes to screening these complaints for trend analysis. Addressing this gap, we introduce a semi-parametric topic model named the hierarchically dual Pitman-Yor process (HDPYP). The HDPYP is designed to automatically process and analyze vast volumes of consumer complaints alongside their associated recall statements. The HDPYP not only extracts defect-related topics but also predicts the significance of each consumer complaint and forecasts the topic distribution of subsequent recall statements. We apply the HDPYP using consumer complaint datasets and vehicle recall data from the U.S. automobile sector. Our findings demonstrate the value of the HDPYP to aid firms and regulators in crucial decision-making processes, such as pinpointing pivotal consumer complaints warranting further examination (or those deemed “summary-worthy” in sub-sequent recall statements), identifying product defects, and forecasting recall occurrences in advance. Furthermore, by integrating the outputs of the HDPYP with Large Language Models (LLMs), regulators can efficiently and effectively review and authenticate the recall statements submitted by firms.