MLentory Knowledge Graph for AI models

MLentory (https://mlentory.zbmed.de/) is centered around information on ML models, how to harmonize that data, and how to make it available and searchable on an FDO (FAIR Digital Object) registry. Our goal is to build a system that extracts ML (Machine Learning) model information from different platforms, normalizes that data in a common format, stores it, and shares it in a FDO registry to facilitate Information Retrieval and comparison/recommendation systems. Currently MLentory aggregates data from Hugging Face, OpenML [1], and AI4Life/Bioimage Model Zoo and provides it as an RDF Knowledge Graph using schema.org [2], Croissant ML [3], CodeMeta [4], and FAIR4ML [5]. MLentory registry on the web uses W3ID identifiers and provides machine-actionable data in the form of Webby FDOs [6], using RO-Crate [7] and FAIR Signposting [8]. This MLentory dataset correspondes to a pre-beta release including the following files: Turtle sample data for models from AI4Life, OpenML, and Hugging Face - mlentory_sample_data_AI4Life_OpenML_HF.turtle JSON-LD sample data for models from AI4Life, OpenML, and Hugging Face - mlentory_sample_data_AI4Life_OpenML_HF.json-ld Preliminary data for models from Hugging Face - 100K_HF.tar.gz As a pre-beta release, this data may need improvements and may not be complete (i.e., it does not contain all the models from all the aggregated platforms). With this early release, we aim at providing a sneek-preview of what will become periodical releases of MLentory. Feedback can be provided as GitHub issues at https://github.com/zbmed-semtec/mlentory-etl-pipeline. A presentation about MLentory is available [9], as well as its Technical Design Document [10]. Funding MLentory FDO registry is part of the NFDI4DataScience consortium, funded by the Deutsche Forschungsgemeinschaft (DFG —German Research Foundation) under the grant number 460234259. This work was supported by the de.NBI Cloud within the German Network for Bioinformatics Infrastructure (de.NBI) and ELIXIR-DE (Forschungszentrum Jülich and W-de.NBI-001, W-de.NBI-004, W-de.NBI-008, W-de.NBI-010, W-de.NBI-013, W-de.NBI-014, W-de.NBI-016, W-de.NBI-022). References Vanschoren J, van Rijn JN, Bischl B, Torgo L. OpenML: networked science in machine learning. SIGKDD Explor Newsl. 2014;15: 49–60. doi:10.1145/2641190.2641198 Guha RV, Brickley D, Macbeth S. Schema.org: evolution of structured data on the web. Commun ACM. 2016;59: 44–51. doi:10.1145/2844544 Akhtar M, Benjelloun O, Conforti C, Gijsbers P, Giner-Miguelez J, Jain N, et al. Croissant: A Metadata Format for ML-Ready Datasets. Proceedings of the Eighth Workshop on Data Management for End-to-End Machine Learning. 2024. pp. 1–6. doi:10.1145/3650203.3663326 Jones MB, Boettiger C, Mayes AC, Arfon Smith, Slaughter P, Niemeyer K, et al. CodeMeta: an exchange schema for software metadata. KNB Data Repository. KNB Data Repository; 2016. doi:10.5063/SCHEMA/CODEMETA-1.0 Castro LJ, Garijo D, Rebholz-Schuhmann D, Ciuciu-Kiss JT, Solanki D, RDA FAIR4ML Metadata Task Force. FAIR4ML metadata schema. 2024. Available: https://rda-fair4ml.github.io/FAIR4ML-schema/release/0.0.1/index.html Soiland-Reyes S, Sefton P, Leo S, Castro LJ, Weiland C, Sompel HV de. Practical webby FDOs With RO-Crate and FAIR Signposting: Experiences and Lessons Learned. Open Conference Proceedings. 2024. doi:10.52825/ocp.v5i.1273 Soiland-Reyes S, Sefton P, Crosas M, Castro LJ, Coppens F, Fernández JM, et al. Packaging research artefacts with RO-Crate. Data Science. 2022; 1–42. doi:10.3233/DS-210053 Van de Sompel H. FAIR Digital Objects and FAIR Signposting. 2023 May 27. doi:10.5281/zenodo.7977333 Quiñones Virgen ND, Rebholz-Schuhmann D, Castro LJ. MLentory: A Machine Learning model registry with natural language queries. 5th conference for Research Software Engineering in Germany (deRSE25); Zenodo; 2025 Mar 6; Karlsruhe. doi: 10.5281/zenodo.14981548. Quiñones Virgen ND, Castro LJ, “MLentory Technical Design Document”. Zenodo, May 26, 2025. doi: 10.5281/zenodo.16943739.