In this study, the shrinking-society risk index (SSRI), utilizing unstructured textual big data to detect and quantify the shrinking-society crisis facing coastal regions in South Korea, is proposed. Recognizing that traditional indicators based on demographic statistics have limitations in the immediate measurement of latent functional degradation within communities, large-scale textual data are defined as environmental and social input signals and a marine soft-sensor framework for multidimensional analysis is designed. The architecture of this framework curated 15 years of news big data and municipal policy documents to perform signal refinement, and was configured to measure social functions through six core parameters: population, economy, healthcare, education, policy response, and social capital. Specifically, in the transducer domain for index derivation, the framework integrates cosine similarity and sentiment analysis algorithms to quantify semantic distances between documents, thereby yielding the functional risk levels of each coastal municipality as quantitative output signals. The experimental results revealed a structural discrepancy between conventional demographic-based risk assessments and the functional risk levels measured by the SSRI. Particularly in the coastal regions of Kangwon province, it was confirmed that the policy discourse related to fishery-based livelihoods acts as a buffering signal that delays the intensification of the shrinking-society crisis. These findings suggest that the diagnostic paradigm for shrinking societies must transition from a uniform demographic-centric model to a multidimensional soft-sensing system focused on functional sustainability. This study provide a technical foundation for intelligent monitoring tools in public policy formulation through the presented sophisticated model that diagnoses region-specific vulnerabilities and administrative responsiveness.