LLMs4Subjects - LLM-based Subject Tagging for the TIB Technical Library’s Open-Access Catalog

NFDI4DS partners organize a Shared Task at GermEval 2025 co-located with Konvens 2025.

We believe libraries are at the heart of our society and education, therefore we believe in impacting technological innovation in these age-old institutions as well as modern digital libraries when we can! We believe in transforming workflows in modern digital library systems. We believe AI can make things faster, we also believe AI can be built to be reliable.

The way we propose to transform workflows in modern digital libraries is to start with one of the core tasks in libraries, i.e. subject indexing. Subject specialists at libraries currently undergo a tedious manual process of memorizing large subject taxonomies and then figuring out a way to ensure that each record submitted to the library has a good coverage of its representative subject tags. This takes a lot of time and effort. Instead, bringing in AI, especially modern technologies like LLMs, to the table can make the process smart, simple, and user-friendly. This goes a long way in making a mark of technological innovation in society.

This is why we created the LLMs4Subjects shared task to provide the opportunity in the community to come up with many innovative solutions around using AI i.e. LLMs for subject indexing. We give you a large-scale annotated dataset, compiled over many years, to build your systems. What we need are creative ideas to this problem. Would you build a RAG approach, or finetune a model, or maybe innovate around few-shot or chain-of-thought prompting?

The 2nd LLMs4Subjects shared task highlights the importance of efficiency in LLMs, encouraging participants to explore strategies that enhance model performance while optimizing for energy consumption and inference speed. We welcome approaches (but not limited to) that leverage model compression, quantization, efficient fine-tuning, and adaptive computation techniques to push the boundaries of sustainable AI development.

Find detailed information on the task page: https://sites.google.com/view/llms4subjects-germeval/home.

For the previous edition see: https://sites.google.com/view/llms4subjects/home/.