What is it about?

This report summarizes the first BIRDS 2025 workshop, held at TPDL 2025, which brought together digital library practitioners and computer science researchers for a full-day program of keynote, panel, and breakout discussions. The workshop addressed how digital libraries should adapt in the era of large language models, covering topics such as retrieval-augmented generation, automated summarization and classification of electronic theses and dissertations, and evaluation methodology for AI-generated outputs. Discussions also examined the tension between research prototypes and production-ready systems, and the challenge of designing user-centered digital library systems when users are often unaware of what current technology can offer. The consensus was that effective progress requires sustained collaboration between researchers, developers, and domain experts.

Featured Image

Why is it important?

This workshop report captures a structured dialogue between digital library practitioners and computer science researchers at a time when large language models are actively reshaping how people access and interact with scholarly content. It addresses practical, unresolved questions: how to evaluate AI-generated summaries and subject classifications of academic documents, how much automation is appropriate versus domain expert involvement, and what it takes to move research systems into production. These questions are relevant across digital libraries, archives, and any institution managing large document collections. The report consolidates current thinking from both communities and identifies concrete challenges that the field needs to work through as AI tools become more embedded in library infrastructure.

Perspectives

Honestly, I think AI gets a bad reputation in the library world — people worry it will replace librarians or water down carefully curated collections. But from what I saw at this workshop, the more pressing issue is that we are not yet using it effectively enough. There are people managing enormous collections of dissertations, archives, and scholarly documents with limited resources, and AI could genuinely take a lot of that burden off their hands. The challenge is not the technology itself but figuring out how to evaluate it, when to trust it, and when to hand back control to a domain expert. What struck me most was how much researchers and practitioners still talk past each other on these questions. Getting those two groups in the same room, working through real problems together, felt like exactly the right approach.

Hajra Klair
Virginia Polytechnic Institute and State University

Read the Original

This page is a summary of: Report on the 1st Workshop on Building Innovative Research Systems for Digital Libraries (BIRDS 2025) at TPDL 2025, ACM SIGIR Forum, December 2025, ACM (Association for Computing Machinery),
DOI: 10.1145/3799914.3799931.
You can read the full text:

Read

Resources

Contributors

The following have contributed to this page