インフラ建設プロジェクトのサステナビリティ最優良管理慣行のためのAIベース環境コード審査ツール

Unofficial AI-generated summary based on the public title and abstract. Not an official translation.

🔬研究者:NLPとLCAを統合した審査自動化の手法と精度検証は、AI×ESG研究の実装例として有用。

🏢実務担当者:建設・インフラ企業は、Envision等の評価文書審査をAIで効率化し、開示・認証対応コストを削減できる可能性がある。

🏛政策担当者:公共インフラ調達や環境規制の適合確認にAI審査を組み込む際の、精度・スケーラビリティの根拠を提供する。

Infrastructure construction and operation are major contributors to global greenhouse gas emissions, which account for approximately one-third of global CO₂ emissions and add to long-term environmental and air quality challenges. Sustainability rating systems such as Envision for infrastructure construction projects provide structured guidance to help teams reduce these impacts, but verifying compliance requires reviewing large volumes of project documents by hand, making the process time consuming and difficult to scale. Therefore, this research develops an AI-based environmental code checking tool that automates the interpretation of project documents and evaluates compliance on lifecycle GHG assessment under Credit CR1.2 of Envision—in other words, determining whether a project successfully reduces greenhouse gas emissions throughout its lifecycle. This research integrates natural language processing, life-cycle assessment, and Envision-based scoring into a unified analytical pipeline. Using a custom spaCy-based Named Entity Recognition (NER) model, the authors trained the system on a representative case study based on real-world empirical data of infrastructure to extract project metadata, material quantities, and activity data from PDF documents. Extracted values were standardized and processed through an LCA module to calculate total, annualized, and intensity-based emissions, which were then benchmarked against baseline scenarios to determine Envision performance levels. A Streamlit web application was created to enable rapid document review with automated extraction, emissions calculation, visualization, and credit scoring. Based on a bridge case study, the system achieved high extraction accuracy with an F1-score of 95.6%, completed assessments in under five seconds, and correctly classified the project as “Improved” with a 17.5% GHG reduction. These findings demonstrate that AI can significantly reduce the time and effort required to evaluate sustainability performance while improving consistency and scalability, supporting broader adoption of standardized digital documentation, early-stage LCA integration, and AI-assisted verification practices in infrastructure construction projects

• semanticscholar https://doi.org/10.31979/mti.2026.2524first seen 2026-09-26 05:30:56 · last seen 2026-09-29 05:41:27