LibraryAI - Project details

Exploratory Development and Testing of an Application for a Virtual Library Research Assistant

LibraryAI is a collaborative project funded by the German Research Foundation (DFG) of the SUB Göttingen, the research group for Scientific Information Analytics (GippLab) at the Institute of Computer Science at the University of Göttingen, and the GWDG. The project utilizes Large Language Models (LLMs) to optimize the use of scientific resources and services at the library, as well as the research opportunities and overall user experience for researchers.
To achieve this, a virtual library assistant is being developed to assist researchers in interacting with the library's scientific resources. Through an interactive chat interface, users can retrieve information, ask questions, request summaries, and translate content. By enabling conversations about the collection material, the virtual assistant will provide dynamic and interactive support for researchers.
To realize these goals, the project uses Transfer Learning to adapt open-source LLMs to the specific needs and terminologies of the library. To address challenges with LLMs, such as outdated information or hallucinations, the virtual assistant will be integrated into existing information systems and library APIs. LibraryAI features a modular architecture that operates independently of underlying library systems and collections. This open and adaptable solution aims to make LibraryAI usable for other libraries as well, ensuring broad adaptability.

Affiliates

Project board at the SUB Göttingen

SUB Göttingen departments / units involved in the project