Duplicate Question Detection in Stack Overflow Based on Extracting Semantic and Statistical Features by Considering Question Score

From a control systems perspective, large‐scale Community Question Answering (CQA) platforms, such as Stack Overflow, can be conceptualized as complex sociotechnical systems in which efficiency, stability, and robustness emerge from continuous interactions between users and an evolving knowledge base. The performance of such systems critically depends on the timely and accurate flow of information, as well as on mechanisms that regulate user behavior and content dynamics. A major challenge in this setting is the prevalence of duplicate questions, which introduce redundancy and informational noise, inflate the workload of moderators and contributors, and fragment valuable knowledge across multiple parallel threads. These effects can be viewed as disturbances that degrade system performance and impair the reliability of information retrieval. Accordingly, the development of automated methods for detecting and managing duplicate questions can be interpreted as the design of feedback control strategies aimed at improving overall system behavior. By identifying redundancy and redirecting user queries toward existing high‐quality content, duplicate detection acts as a control input that mitigates inefficiencies, stabilizes information flow, and promotes optimal utilization of human and computational resources. In this framework, duplicate question detection is not merely a classification task but a regulatory mechanism that preserves the structural integrity of the knowledge base, supports consistent content quality, and prevents the dispersion of expert insights. Framing the problem through the lens of control theory provides a principled foundation for analyzing platform dynamics and for designing intelligent moderation tools that enhance the scalability and long‐term sustainability of CQA ecosystems.