Applications of LLM and NLP in the Retrieval and Analysis of Institutional Publications

Abstract

Modern higher education institutions are increasingly requiring advanced tools to understand and analyze their research output effectively. This paper explores the integration of Large Language Models (LLMs) and Natural Language Processing (NLP) to develop an advanced system for institutional publication analysis at the University of Münster. We demonstrate how Retrieval Augmented Generation (RAG) pipelines can enhance the discovery and analysis of research documents through semantic search and intelligent information processing. Our current approach integrates vector-based document representation with graph-based modeling to construct a RAG pipeline for publication retrieval and analysis. Although still in early development, the system is designed to support several use cases, including the identification of research trends and collaboration patterns, as well as the generation of summaries and reports for specific research topics. We share practical insights from the implementation of this system and discuss technical solutions to common challenges related to the processing and analysis of scientific publications.