한국 물류산업의 ESG 담론 구조 분석: TF-IDF와 SBERT 기반 뉴스 데이터 마이닝 접근

초록

This study maps initiative-level Environmental, Social, and Governance (ESG) themes in Korea’s logistics sector using a corpus of 837 news articles spanning from 2011 to September 2025. Themes are derived using Term Frequency?Inverse Document Frequency (TF-IDF)?guided vocabulary selection and by clustering Sentence-BERT (SBERT) embeddings with a bisecting k-means tree; cohesion?separation diagnostics identify K=16 as the baseline that best represents the cluster structure, with K=32 used for higher-resolution partitions. The annual series shows a clear post-2020 inflection, peaking in 2021 at 132 articles and then moderating to 102 in 2022 and 72 in 2023. Coverage is heavily weighted toward environmental topics (63.3%), with social?labor secondary (21.0%) and governance minimal (0.4%), indicating an agenda imbalance. Methodologically, the TF-IDF→SBERT→tree pipeline supports consistent labeling and longitudinal tracking across logistics subfields. The research yields interpretable, initiative-level themes aligned with the K=16 structure: green packaging and energy efficiency (e.g., waste-heat recovery), port and yard equipment modernization (e.g., yard tractors and transfer cranes), domestic logistics nodes, capacity and chartering, labor and operations regulation, public-sector and municipal programs, and a general named-entity background class used for filtering. Therefore, the study delivers a reproducible map of initiative-level ESG themes and a practical toolkit for agenda tracking in logistics media.