CSRMatchNet: An intelligent semantic decision support system for evaluating firm-customer CSR congruence

Corporate social responsibility (CSR) effectiveness depends not only on disclosure volume but also on whether firms address the CSR themes that customers genuinely care about. Existing CSR assessments primarily rely on aggregated external ratings or firm-reported information, which obscures firm-customer misalignment at the level of specific CSR themes and provides limited actionable guidance. This study proposes CSRMatchNet, an intelligent semantic decision support system that evaluates firm-customer CSR congruence at the level of individual CSR themes. CSRMatchNet directly derives CSR themes from large-scale firm-authored CSR disclosures and customer-generated CSR discourse, without relying on predefined taxonomies or external ratings. The system embeds firm and customer texts into a shared semantic space using a pretrained multilingual Sentence-BERT model and applies a lightweight residual adapter trained with covariance alignment (CORAL) and multi-kernel maximum mean discrepancy (MMD) to align cross-source distributions. Unsupervised clustering is used to identify recurring CSR themes, and theme-level firm-customer alignment is quantified based on semantic divergence between firm and customer representations. To enhance decision relevance, alignment scores are weighted by customer salience to construct an Effective CSR (ECSR) measure, capturing the realized market-facing effectiveness of CSR efforts rather than disclosure intensity alone. An empirical application to major Chinese firms demonstrates that CSRMatchNet produces fine-grained CSR diagnostics that distinguish shared priorities, CSR gaps, over-communicated themes, and low-priority themes. External validation is performed using the CSR Market Performance Index (CMPI) to evaluate whether firm-level ECSR signals exhibit meaningful external relevance and interpretability. Overall, CSRMatchNet advances CSR analytics by providing an AI-enabled, transparent, and actionable decision support framework for aligning CSR strategies with customer expectations.