Data-driven lexical framing of ESG in infrastructure projects: an NLP-based decision support approach

Purpose This study investigates how environmental, social and governance (ESG) priorities and digital transformation are linguistically framed in project-level discourse within sustainable infrastructure delivery. It examines how language shapes managerial attention, governance priorities and the interpretation of sustainability evidence, beyond formal indicators and dashboards. By applying natural language processing (NLP) to large-scale practitioner communications, the study aims to demonstrate how discourse-aware analytics can enhance transparency, reflexivity and balance in ESG-related decision-making and support more socially sensitive and context-aware project governance. Design/methodology/approach The study adopts a data-driven design using a large corpus of infrastructure project communications. A hybrid NLP pipeline is applied, combining topic modelling, semantic clustering and sentiment analysis to identify dominant ESG themes and evaluative patterns. To ensure interpretability and contextual validity, automated outputs are iteratively validated through human-in-the-loop expert review. The approach focuses on project-level discourse, enabling systematic analysis of how ESG issues, as well as digital technologies, are framed in practice to inform decision support and governance design. Findings The analysis reveals a pronounced asymmetry in ESG discourse within infrastructure projects. Environmental and governance themes are frequently framed in positive, data-driven terms and closely linked to digital tools such as dashboards, automation and monitoring systems. In contrast, social issues are less visible, more cautiously articulated and rarely digitally mediated, reflecting relational and normative framing. Sentiment analysis shows optimism associated with digitalized ESG narratives, suggesting a digitization bias that amplifies measurable dimensions while marginalizing social concerns. These findings indicate that language and digital mediation jointly shape managerial attention and may reproduce imbalances in ESG-oriented decision support. Research limitations/implications The study is limited by its sectoral focus on infrastructure projects, which may constrain generalizability to other industries. The use of lexicon-based sentiment analysis and unsupervised topic modelling may oversimplify nuanced or context-dependent ESG language, while potentially introducing algorithmic bias. In addition, the cross-sectional design captures a static snapshot of discourse rather than temporal evolution. Despite human-in-the-loop validation, some interpretive subjectivity remains in framing classification. Future research could extend this work through longitudinal designs, domain-specific NLP models and cross-sector validation to strengthen the robustness and generalizability of findings. Practical implications The study offers implications for infrastructure project governance and decision-support system (DSS) design. It demonstrates how ESG-related communication can be translated into actionable framing indicators, enabling more transparent and reflexive decision-making. Practically, organizations can integrate framing analytics into dashboards to complement traditional ESG metrics by visualizing sentiment, topic salience and digital framing patterns. This supports early detection of imbalances, such as the under-representation of social sustainability or over-reliance on digital optimism narratives. However, these tools should be used as decision-support aids rather than automated decision-making mechanisms, given the risk of misinterpretation of contextual and linguistic signals. Originality/value This study advances sustainable infrastructure research by introducing lexical framing as a form of “decision-metadata” that reveals how ESG priorities are interpreted and legitimized in project governance. Unlike prior ESG analytics that focus on indicators or predictive accuracy, it integrates NLP with interpretive, human-in-the-loop analysis to diagnose rhetorical asymmetries across ESG pillars. The study demonstrates how digital transformation shapes not only ESG measurement but also sustainability narratives, offering a novel, discourse-aware approach to decision support that enhances transparency, reflexivity and social balance in infrastructure project governance.