No Dedicated IT Team, No Limits: Automating Research Output Management with AI at ETRI Library

Abstract

Accumulating research outputs in an institutional repository does not guarantee discoverability. The Electronics and Telecommunications Research Institute (ETRI) Library, part of Korea's largest government-funded ICT research institute, addressed the low visibility of its repository records in Google Scholar as a metadata quality problem rather than a search engine optimization issue. Using AI-assisted coding, the library developed scripts to inspect metadata and monitor indexing status, which improved Google Scholar indexing. Missing abstracts emerged as a major challenge, so for papers whose abstracts were unavailable from external databases, the library created a workflow combining OCR and a large language model (LLM) to identify abstract sections directly from original PDFs. Building on this work, the library is expanding automation through Codex-based agents and robotic process automation (RPA) to collect official conference and journal URLs and support the entry and verification of publication records. This case shows that libraries can achieve meaningful innovation without dedicated IT specialists or large budgets by improving metadata, automating repetitive tasks, and keeping final validation in human hands.