ESG-KG: A Multi-modal Knowledge Graph System for Automated Compliance Assessment
ESG-KG: 自動コンプライアンス評価のためのマルチモーダル知識グラフシステム (AI 翻訳)
Li-Yang Chang, Chih-Ming Chen, Hen-Hsen Huang, Ming-Feng Tsai, An-Zi Yen, Chuan-Ju Wang
🤖 gxceed AI 要約
日本語
ESG-KGは、企業のサステナビリティ報告書からテキスト、表、図、インフォグラフィック等のマルチモーダル情報を抽出し、知識グラフを構築するシステムである。LLMとRAGを統合し、ESGフレームワークへの準拠を自動評価する。エンドツーエンドのパイプラインをデモで示し、解釈可能な評価結果と証拠連鎖を提供する。
English
ESG-KG is a system that extracts multi-modal information (text, tables, figures, infographics) from corporate sustainability reports to build a knowledge graph. It integrates LLM-based reasoning with retrieval-augmented generation to automatically assess compliance with ESG frameworks. The demonstration showcases an end-to-end pipeline producing interpretable results with traceable evidence chains.
Unofficial AI-generated summary based on the public title and abstract. Not an official translation.
📝 gxceed 編集解説 — Why this matters
日本のGX文脈において
日本ではSSBJ基準や有価証券報告書でのESG開示が進む中、本システムは開示データの自動評価・検証ツールとして実務に活用できる。特にマルチモーダル対応により、図表や写真を含む統合報告書の分析が可能で、日本の開示実務に即したコンプライアンス評価に貢献する。
In the global GX context
As global frameworks like TCFD, ISSB, CSRD, and SEC climate rules mandate detailed ESG disclosures, this system offers a scalable method for automated compliance assessment. Its knowledge graph architecture enables cross-modal evidence tracing, which is critical for auditors, investors, and regulators seeking verifiable compliance checks.
👥 読者別の含意
🔬研究者:Novel integration of KG and LLM for automated ESG compliance; open-source toolkit facilitates reproducibility and further research.
🏢実務担当者:Provides corporate sustainability teams with a tool to automate compliance scoring against multiple ESG frameworks, reducing manual effort.
🏛政策担当者:Demonstrates feasibility of automated compliance monitoring, potentially informing regulatory technology (RegTech) for ESG disclosure enforcement.
📄 Abstract（原文）
We present ESG-KG, a system that auto-mates ESG compliance assessment through multi-modal information extraction and knowledge graph construction.ESG-KG processes corporate sustainability reports containing diverse data formats—text, tables, figures, and infographics—and extracts ESG-related entities, relationships, and metrics into a structured knowledge graph. This KG-based architecture enables precise cross-modal information retrieval and provides verifiable evidence grounding for downstream analysis.Built upon this foundation, ESG-KG integrates retrieval-augmented generation (RAG) with LLM-based reasoning to automatically evaluate compliance against ESG frameworks and standards. Our demonstration show-cases the system’s end-to-end pipeline, from multi-modal document processing to automated compliance scoring, highlighting its capability to handle real-world sustainability reports and generate interpretable assessment re-sults with traceable evidence chains. To facilitate further research, we release our open-source Python toolkit for Automated Compliance Assessment at https://github.com/ cnclabs/website.kg.esg.demo.git , and a live demonstration video is available at https: //youtu.be/Lj4Zp74J1nY .
🔗 Provenance — このレコードを発見したソース
- semanticscholar https://doi.org/10.18653/v1/2026.eacl-demo.43first seen 2026-05-15 18:54:29
🔔 こうした論文の新着を逃したくない方は キーワードアラート に登録（無料・3キーワードまで）。
gxceed は公開メタデータに基づく研究支援データセットです。要約・翻訳・解説は AI 支援で生成されています。 最終的な解釈・検証は利用者が原典資料に基づいて行うことを前提とします。