A system-level semantic reliability framework for diagnosing governance failures in high-risk socio-technical systems

High-risk industries represent complex socio-technical systems in which governance failures often arise from the interaction of technical, organizational, and cultural factors. Traditional reliability approaches rely predominantly on accident records or audits, which provide limited insight into governance and safety culture dimensions. To address this limitation, this study develops a semantic diagnostic framework that interprets corporate social responsibility(CSR) reports as indicators of governance performance. Based on 2786 reports issued by 372 enterprises between 2010 and 2023, topic modeling was applied to identify latent governance modules, which were then tracked across regulatory phases using a lifecycle-based Stable–Transformative–Dissolving–Emerging(STDE) typology. Network-based and multivariate analyses were applied to examine governance maturity, coupling, and structural change patterns. Results indicate a sectoral transition from reactive compliance toward more integrated governance structures. Core functions such as hazard identification and employee training remain stable, while investment-oriented themes decline and new governance emphases emerge. Cross-industry comparison indicates relatively mature governance patterns in mining, transitional trajectories in construction and manufacturing, and greater structural diversity in transportation. Overall, the framework provides a complementary, governance-oriented perspective for longitudinal monitoring and cross-sector comparison of safety-related practices, without relying on direct accident or failure data.