Text Mining and Machine Learning for New-MediaPublic-Opinion Spillovers in Manufacturing Supply-Chain Digital Collaboration

Online media coverage may transmit information across supply-chain relationships, yet converting such signals into evidence requires a clear separation between text measurement, network alignment, and statistical inference. This study examines whether focal-firm media attention is associated with connected partners’ annual-report digital disclosure and develops a manufacturing-oriented workflow for directed supply-chain analysis. We construct a firm–partner–year panel for Chinese non-financial A-share firms from 2007 to 2024 by combining aggregated CFND media indicators, annual-report digital-term frequencies, CNRDS supplier–customer links, and CSMAR controls. Fixed-effects models estimate conditional associations on directed edges. The baseline specification shows a positive association between focal-firm coverage and partner digital disclosure (β = 0.048, SE = 0.020), while alternative aggregate coverage measures yield coefficients with the same sign; mechanism and subgroup results remain observational and descriptive. These findings do not establish causality or realized interfirm digital collaboration because annual-report word frequency is an imperfect disclosure proxy. A Chinese- BERT–MLP event-and-tone module is therefore retained only as a future extension pending article-level text, human- reviewed labels, and out-of-sample evaluation.