앙상블 기반 OCR과 RAG 기술을 활용한 ESG 서류 검토 자동화 시스템 개발
Development of an Automated ESG Document Review System using Ensemble-Based OCR and RAG Technologies
1
조회
0
다운로드
0
인용
0
참고문헌
키워드
초록
본 연구는 ESG 서류 검토 과정의 효율성 향상을 위해, 광학 문자 인식(OCR)과 검색 증강 생성(RAG) 기술을 융합한 새로운 자동화 시스템을 제안한다. 제안된 시스템은 OCR 프로세스에 앙상블 모델 기반의 이미지 전처리 알고리즘과 하이브리드 정보 추출 모델을 적용하여 텍스트 인식의정확도를 향상시키며, RAG 파이프라인에 레이아웃 분석 알고리즘과 재순위화 알고리즘, 앙상블검색기 등을 적용하여 정보 검색과 답변 생성의 신뢰성을 최적화한다. 시스템의 성능을 평가하기위해 온라인 포털에 게시된 인증서 이미지와 기업 웹사이트 등에 공개된 회사 내규를 사용하여테스트를 진행한 결과, 인증서 검토에서 93.8%, 회사 내규 검토에서 92.2%의 정확도를 달성하며, 제안된 시스템이 ESG 평가 과정에서 인간 평가자를 효과적으로 보조할 수 있음을 보여주었다.
Abstract
This study proposes a novel automation system that integrates Optical Character Recognition (OCR) and Retrieval-Augmented Generation (RAG) technologies to enhance the efficiency of the ESG (Environmental, Social, and Governance) document review process. The proposed system improves text recognition accuracy by applying an ensemble model-based image preprocessing algorithm and hybrid information extraction models in the OCR process. Additionally, the RAG pipeline optimizes information retrieval and answer generation reliability through the implementation of layout analysis algorithms, re-ranking algorithms, and ensemble retrievers. The system's performance was evaluated using certificate images from online portals and corporate internal regulations obtained from various sources, such as the company’s websites. The results demonstrated an accuracy of 93.8% for certification reviews and 92.2% for company regulations reviews, indicating that the proposed system effectively supports human evaluators in the ESG assessment process.
참고문헌 (17)
- Raj, Aaryan, "Revolutionizing data entry : An in-depth study of optical character recognition technology and its future potential", International Journal for Research in Applied Science &Engineering Technology, 11(2), pp.645~653 (2023)[DOI]
- Lewis, Patrick, "Retrieval-augmented generation for knowledge-intensive nlp tasks", Advances in Neural Information Processing Systems, 33, pp.9459~9474 (2020)
- Ori Ram, "In-Context Retrieval-Augmented Language Models", Transactions of the Association for Computational Linguistics, 11, pp.1316~1331 (2023)[DOI]
- Guu, Kelvin, "Retrieval augmented language model pre-training", International conference on machine learning, pp.3929~3938 (2020)
- Indyk, Piotr, "Approximate nearest neighbors : towards removing the curse of dimensionality", Proceedings of the thirtieth annual ACM symposium on Theory of computing, pp.604~613 (1998)
- Ye, Hongbin, "Ontology-enhanced Prompt-tuning for Few-shot Learning", Proceedings of the ACM Web Conference (2022)
- "LangChain Introduction"[URL]
- "LlamaIndex"[URL]
- Liu, Nelson F., "Lost in the middle : How language models use long contexts", Transactions of the Association for Computational Linguistics, 12, pp.157~173 (2024)[DOI]
- Bui, Quang Anh, "Selecting Automatically Pre-Processing Methods to Improve OCR Performances", 2017 14th IAPR International Conference on Document Analysis and Recognition (ICDAR), pp.169~174 (2017)
- Cheng, H. D., "Detecting of contrast over-enhancement", 2012 19th IEEE International Conference on Image Processing, pp.961~964 (2012)
- "Ensemble Retirever"[URL]
- Ouyang, Long, "Training language models to follow instructions with human feedback", Advances in neural information processing systems, 35, pp.27730~27744 (2022)
- "Public Institution Disclosure Site"[URL]