The Next Generation of Data Management with Artificial Intelligence

TY - GEN

T1 - The Next Generation of Data Management with Artificial Intelligence

AU - Tykhonov, Vyacheslav

AU - Rijsselberg, Fjodor van

AU - Endarto, Eko

PY - 2024/12/17

Y1 - 2024/12/17

N2 - Building a sustainable and distributed research data infrastructure for the future, utilizing automated metadata translation and enrichment through the integration of knowledge graphs with Large Language Models in the ODISSEI project. An important obstacle with data-driven research lies with the difficulty of knowing what questions the data might have the answers to. Even if rich metadata is provided, these alone are often not sufficient to guide researchers towards new lines of research, which may leave many potentially interesting and novel insights to remain undiscovered. This problem has only exacerbated with the recent digitalization efforts of paper collections and libraries, placing vast amounts of cross-domain heterogeneous data at the researcher’s fingertips. Many of these datasets are now also being published as knowledge graphs, following Linked Data principles, thereby integrating multiple relevant sources of data that have never been together before. This provides unique opportunities for pattern detection methods, which can now leverage the graphs’ deeper relational dependencies and semantics to discover and highlight potentially interesting correlations in the data. Researchers can then inspect these correlations, and, if deemed relevant, use them as starting points for new research directions or as support for existing lines of research.

AB - Building a sustainable and distributed research data infrastructure for the future, utilizing automated metadata translation and enrichment through the integration of knowledge graphs with Large Language Models in the ODISSEI project. An important obstacle with data-driven research lies with the difficulty of knowing what questions the data might have the answers to. Even if rich metadata is provided, these alone are often not sufficient to guide researchers towards new lines of research, which may leave many potentially interesting and novel insights to remain undiscovered. This problem has only exacerbated with the recent digitalization efforts of paper collections and libraries, placing vast amounts of cross-domain heterogeneous data at the researcher’s fingertips. Many of these datasets are now also being published as knowledge graphs, following Linked Data principles, thereby integrating multiple relevant sources of data that have never been together before. This provides unique opportunities for pattern detection methods, which can now leverage the graphs’ deeper relational dependencies and semantics to discover and highlight potentially interesting correlations in the data. Researchers can then inspect these correlations, and, if deemed relevant, use them as starting points for new research directions or as support for existing lines of research.

U2 - 10.5281/zenodo.14507120

DO - 10.5281/zenodo.14507120

M3 - Other contribution

ER -