LINAC2026 - 33rd Linear Accelerator Conference

Performance Comparison of TeX Preprocessing Methods and Embedding Models for RAG on RAON Control System Technical Notes

Not scheduled

2h

Daejeon Convention Center

Poster Presentation MC4.A05: Other technology Poster Session

Speaker

Won Hee Min (Chungnam National University, Institute for Basic Science)

Description

RAON uses EPICS to integrate local control systems into central control. Technical notes for system installation and management are documented in TeX format and managed with relevant code and files in a Git-based configuration management environment on an internal network. This study examines the applicability of Retrieval-Augmented Generation (RAG) to control system technical documents for the introduction of AI technologies in control infrastructure.

TeX documents were converted into formats such as Markdown under multiple preprocessing scenarios with different metadata augmentation and chunk segmentation conditions. Different embedding models were applied to construct vector databases. Relevant queries were designed, and performance was compared in terms of retrieval accuracy and contextual relevance. This study aims to evaluate the feasibility of applying RAG to EPICS technical documents and to suggest directions for building a retrieval framework for TeX-based documentation.