scieee AI-readable full text Open interactive document viewer

PaN-Finder

Novelli, Massimiliano

Abstract

The PaN-Finder project aims to enhance the capabilities of the existing PaNOSC Data Portal, which provides researchers access to a wealth of data from Europe’s Photon and Neutron (PaN) Research Infrastructures (RIs). Originally developed as part of the EU Horizon 2020 project PaNOSC, the portal has already improved the visibility and accessibility of scientific data. PaN-Finder will introduce an AI-powered search tool that simplifies user interaction, enabling intelligent, prompt-based searches. By leveraging state-of-the-art AI technologies, such as Large Language Models and Natural Language Processing, this tool will make it easier for researchers, journalists, and the general public to navigate and utilise valuable PaN data effectively. This talk was held in the course of the DAPHNE4NFDI TA1 Data for science lecture series on October, 28 2025

Full text

PaN-Finder Max Novelli and the SIMS team (DMSC - ESS) DAPHNE4NFDI lecture series, 2025-10-28 2025-10-28 DAPHNE4NFDI LECTURE SERIES - 2025/10/28 2 The beginning Greg Razoky on Unsplash DAPHNE4NFDI LECTURE SERIES - 2025/10/28 3 PaNOSC Data Portal DAPHNE4NFDI LECTURE SERIES - 2025/10/28 4 Success!!! Data Portal Federated Search PaNOSC Search Data catalog ESS Facilities Infrastructure PaNOSC Infrastructure ESRF ILL Max IV PSI DAPHNE4NFDI LECTURE SERIES - 2025/10/28 5 Few Challenges ●The project ended. Supported as in-kind contribution ●Connection between federated and local search APIs is not robust ●Local search APIs are hard to monitor ●Data curation at each facility varies a lot DAPHNE4NFDI LECTURE SERIES - 2025/10/28 6 User Experience DAPHNE4NFDI LECTURE SERIES - 2025/10/28 7 The Vision DAPHNE4NFDI LECTURE SERIES - 2025/10/28 8 The requirements Centralized Natural Language Artificial Intelligence Retrieval Augmented Generation Asynchronous data collection DAPHNE4NFDI LECTURE SERIES - 2025/10/28 9 The idea I’m writing a new proposal for a new experiment to study the changes in molecular structure of a new synthetic carbon based material bonding with human blood red cells. I’m interested in running the experiments at a normal human body temperature. Can you show me related publications and similar proposals? Have similar experiments already been conducted? In-vivo experiments with synthetic carbon based materials has been already performed and you can find information about them in the following resources: ●Resource 1 (URL 1) ●Resource 2 (URL 2) I have not found information regarding new synthetic carbon based materials bonding with human blood red cells, but relevant information might be found in the following resources: ●Resource 3 (URL 3) ●Resource 4 (URL 4) You also might want to check the following open datasets as they have been acquired from human blood red cells at normal body temperature: ●Dataset 1, RI, DOI ●Dataset 2, RI, DOI ●Dataset 3, RI, DOI DAPHNE4NFDI LECTURE SERIES - 2025/10/28 16 User Prompt Components DAPHNE4NFDI LECTURE SERIES - 2025/10/28 17 Demo EOSC BUILD-UP GROUP MEETING #7 - 2025/10/02 18 Query Examples Query #1 Find datasets where lungs from a covid patient where scanned DOI: 10.15151/ESRF-ES-436648953 Query #2: Can you find me documents where elephants are equal to apples DOI: ????? Bonus Query: 1) Find reflectivity datasets on small samples, surface < 1cm2 2) Find reflectivity datasets on small samples with surface < 1cm2 3) Find reflectometry datasets on small samples with surface < 1cm2 4) Find datasets on small samples with surface < 1cm2 acquired with reflectometry technique DOI: 10.15151/ESRF-ES-1824853946 EOSC BUILD-UP GROUP MEETING #7 - 2025/10/02 19 Query #1: prompt EOSC BUILD-UP GROUP MEETING #7 - 2025/10/02 20 Query #1: Results EOSC BUILD-UP GROUP MEETING #7 - 2025/10/02 21 Query #1: details EOSC BUILD-UP GROUP MEETING #7 - 2025/10/02 22 Query #2: prompt and results DAPHNE4NFDI LECTURE SERIES - 2025/10/28 23 What’s next? DAPHNE4NFDI LECTURE SERIES - 2025/10/28 24 Current and future work ●Benchmarking ●More Testing: ●Different models (size, cost, performance) ●Prompt Engineering ●Model fine tuning ●Refine Natural Language Response ●Automating Data Collection ●Driving better data curation ●User feedback: 1-on-1 sessions, answers validation, and more Thank you for your attention Max Novelli max.no[email protected] Thanks to Janos Babik and the SIMS team Questions?