ScholarWorks@성균관대학교

초록

This study proposes an automated review system that integrates AI-based Optical Character Recognition (AI OCR) and Small Language Models (SLM) to address inefficiencies in the military budget execution review process. While the current electronic review system enhances document management, it also leads to increased review time and a higher risk of human error. The proposed system utilizes AI OCR to automatically extract key information from accounting documents and employs SLM to compare extracted data against budgetary guidelines. Experimental simulations using ChatGPT as a proof-of-concept tool demonstrated the feasibility of this approach for future SLM implementation, showing accurate data extraction and valid review outcomes. The results suggest that the proposed system can improve review efficiency, reduce manpower requirements, minimize errors, and support policy enhancement in military operations. Moreover, it holds potential for broader application across other administrative tasks.

키워드