In this study, we propose a data-driven analytical framework for detecting shrinking-society risk in coastal regions and examining the effect of social innovation activities undertaken by fourth-sector firms on corporate sustainability. To this end, we applied a multi-source data fusion approach that integrates unstructured textual signals with structured financial data. Specifically, domain-specific benchmark lexicons were constructed using natural language processing techniques, including term frequency–inverse document frequency (TF-IDF) and Word2Vec algorithms, and sustainability and social innovation indices were derived from corporate disclosure materials and news texts through cosine similarity analysis. With panel data from 159 coastal-region firms over the period 2016–2025, the empirical results showed that social innovation activities by fourth-sector firms have a significant positive effect on corporate sustainability. The findings further indicated that shrinking-society risk positively moderates this relationship, suggesting that social innovation becomes more important in regions with greater structural vulnerability. This study makes a methodological contribution by transforming social innovation and regional risk into measurable indicators through information processing techniques. It also provides a technical foundation for designing data-driven monitoring systems that support coastal economic self-reliance and evidence-based policy decision-making.