Conference Agenda

This talk will explore how advancements in large language models can help review, write, and improve connections of machine-actionable data management plans. It will review ongoing initiatives at the California Digital Library to improve the benefits and easy burden of writing DMPs, while still keeping the human author as the authority. It will cover learnings and progress and tips for others exploring similar areas.

Researchers find preparing data management plans (DMPs) time consuming and see data stewards as an appealing solution for filling them in on their behalf.

At Eindhoven University of Technology (TU/e), data stewards support in filling DMPs is limited by the knowledge on the research topic.

Advances in technological solutions are perhaps the most feasible way to support researchers in filling their DMPs before the data stewards final approval.

At TU/e different research support groups, such as the Research Data Infrastructure (RDI) and the Data Stewards (DS), have initiated a collaboration with an external party to develop a GenAI tool that would directly and automatically extract the relevant information from research proposals and to pre-fill the DMPs.

The tool is able to extract structured metadata from research proposals, based on a user-defined schema specifying the desired metadata attributes, e.g., research goals, presence of human participants, data origin and size, etc. (including a description per metadata attribute).

This presentation examines how large language models (LLMs) like ChatGPT can assist researchers in generating documentation such as codebooks and README files, helping meet Open Science and FAIR data requirements. It explores how LLMs can streamline workflows, align outputs with repository standards, and reduce curation burden. Alongside these opportunities, the talk also addresses the ethical, legal, and environmental challenges posed by AI integration, especially in contexts involving sensitive social science data.

Integrating diverse data in Earth System Sciences (ESS) is hindered by semantic heterogeneity, as different disciplines use inconsistent terminologies. The BITS project addresses this challenge by providing a dedicated Terminology Service (TS) with over 40 vocabularies, maintained by the TIB for long-term use. A practical use case at the Senckenberg Institute combines AI tools, edge computing, and the TS to automatically annotate and process digitized natural science collections efficiently. This workflow improves data findability and interoperability, particularly for multilingual and handwritten records, while ensuring data security through near-data processing. The Lightning Talk highlights potential risks of irresponsible AI use, despite its clear benefits when integrated with other services.