Integrating OCR and LLMs for Enhanced Document Digitization in ERP Systems
Taehong Lee, Gayeon Kim, Hyungjin Ahn, J. Y. Jeong, Mingyu Jeong, Jiu Song · 2024
Enterprises efficiently manage transactions involving purchases, sales, and human resources through Enterprise Resource Planning (ERP) systems. However, integrating varied formats of analog documents into an ERP system presents significant challenges due to the need for digitization. This research proposes a methodology that combines Optical Character Recognition (OCR) with Large Language Models (LLMs) to digitize and integrate analog documents into the open-source ERP system, ERPNext. By leveraging LLMs, this approach addresses the limitations of systems that rely solely on OCR for document digitization, particularly in cases when OCR struggles with character misrecognition. Consequently, the results confirm that this methodology improves the efficiency and accuracy of corporate document management within ERP systems, ultimately enhancing profitability.