PaperlayerView exact revision on arXiv

Source-linked AI summary

MaRDMO: FAIR Documentation of In-Silico Research

Marco Reidelbach, Marcus Weber

arXiv:2609.11931v1cs.DLcs.SE

TL;DR

Mathematical research data lack broadly adopted management practices and accessible pathways into FAIR infrastructure. MaRDMO embeds mathematics-specific questionnaires and search into RDMO, connecting researchers to the MaRDI ecosystem. Its reuse-first workflow, dedicated instance, simplified model catalog, and class-filtered search make documentation and discovery more practical, while LLM-assisted features remain under development.

  • Problem

    Mathematics is often treated as data-free, leaving research-data-management support without consistent ways to assist mathematicians or apply established standards.

  • Method

    MaRDMO extends RDMO with ontology-driven questionnaires, reuse-first item selection, search-catalog queries, and direct connections to MaRDI knowledge graphs.

  • Results

    MaRDMO provides a practical gateway for documenting and discovering mathematical models, algorithms, and interdisciplinary workflows within an existing RDM environment.

  • Takeaways & Limitations

    Embedding MaRDI functionality in RDMO makes FAIR documentation and knowledge-graph reuse more accessible for everyday research practice.

Abstract

from arXiv · show

MaRDMO is a plugin for the Research Data Management Organiser (RDMO) that enables the structured, FAIR-compliant documentation and discovery of mathematical research data. By embedding mathematics-specific questionnaires into a widely used data management plan tool, MaRDMO lowers the barrier to contributing and querying the MaRDI Knowledge Graph for researchers across all disciplines. This paper presents the current state of MaRDMO, including its questionnaire-driven documentation process for mathematical models, algorithms, and interdisciplinary workflows, illustrated through a concrete example based on a solver comparison study for the Stokes-Darcy system. We further describe the dedicated MaRDI RDMO instance as a ready-to-use entry point for the community, and discuss recent developments including a simplified Basic Model Catalog and an improved class-filtered search. The paper concludes with an outlook on LLM-assisted documentation features currently under development.

1 Introduction

Mathematical research is often treated as data-free, leaving researchers and support staff without consistent data-management practices. MaRDMO addresses this gap by embedding structured, mathematics-specific documentation and discovery into the widely deployed RDMO platform.

  • 1 Introduction: Mathematics is often perceived as data-free because its outputs are expressed as theorems, proofs, and formulas rather than datasets or measurements.This perception leaves research-data-management support staff uncertain how to assist mathematicians, while expectations vary across institutions and funding contexts.
  • 1 Introduction: Applied and computational mathematics generates and transforms data through a model-simulation-optimization cycle spanning inputs, models, algorithms, outputs, validation, and new problems.The workflow’s core research data types are mathematical models, algorithms, and the data they generate and transform.
  • 1 Introduction: MaRDI provides ontologies and knowledge graphs for mathematical models, algorithms, and interdisciplinary workflows as infrastructure for FAIR mathematical research data.These resources include MathModDB and related ecosystem services, but the paper emphasizes that technical infrastructure alone does not ensure adoption.
  • 1 Introduction: MaRDMO extends RDMO with structured questionnaires that document and search mathematical research data without requiring knowledge of SPARQL, Wikibase, or knowledge graphs.The plugin connects existing RDM processes to MaRDI resources and supports mathematical models, algorithms, and interdisciplinary workflows.
  • 1 Introduction: MaRDMO lowers adoption barriers by integrating with RDMO, which is already widely deployed across German research institutions and NFDI consortia.The paper presents the plugin as a way to use existing local RDM processes rather than requiring an entirely new system.

2 The MaRDI Ecosystem

The MaRDI ecosystem combines a knowledge graph with specialized databases and services covering documentation, execution, reproducibility, and reuse. Its interoperability focus makes MaRDMO a practical entry point for contributing to and querying these connected resources.

  • 2 The MaRDI Ecosystem: MaRDI’s knowledge graph links mathematical models, algorithms, workflows, publications, software, proofs, and formulas across the research-data lifecycle.MathModDB and MathAlgoDB provide structured subgraphs whose task links enable cross-type queries between models and algorithms.
  • 2 The MaRDI Ecosystem: The MaRDI Portal is a Wikibase-powered platform hosting a continuously growing, community-maintained knowledge base for mathematical research data.It serves as both a human-readable reference and a machine-queryable resource.
  • 2 The MaRDI Ecosystem: MaRDIFlow supports FAIR computational experiments by designing, executing, and documenting workflows while automatically capturing provenance and metadata.MaPS complements this ecosystem by addressing software reproducibility through unified packaging.
  • 2 The MaRDI Ecosystem: Interoperability keeps mathematical research data linkable across MaRDI services and the broader research-data landscape rather than siloed.MaRDMO provides the practical bridge for researchers who want to document, search, contribute, or reuse models, algorithms, and workflows.

3 The MaRDMO Plugin

MaRDMO extends RDMO with ontology-driven questionnaires that document mathematical models, algorithms, and interdisciplinary workflows, while connecting researchers to the MaRDI Knowledge Graph. Its guided workflow supports reuse of existing metadata, validation, export, and SPARQL-free discovery, illustrated through the Stokes-Darcy solver-comparison example.

  • 3 The MaRDMO Plugin: MaRDMO extends RDMO with mathematics-specific questionnaires and export functionality for documenting models, algorithms, and interdisciplinary workflows in a standardized format.It connects the familiar RDMO interface to the MaRDI Portal and Wikidata and provides a dedicated catalog for querying existing research data.
  • 3 The MaRDMO Plugin: Each supported data type follows its ontology through a search-first workflow that retrieves existing items, pre-fills metadata, and lets researchers extend or create entries when needed.This reduces manual effort while accommodating items distributed across connected knowledge graphs.
  • 3 The MaRDMO Plugin: The model, algorithm, and workflow questionnaires cover complementary research objects, from formulas and computational tasks to software, benchmarks, processing steps, datasets, and hardware.The Stokes-Darcy example documents a coupled fluid–porous-medium model, the Uzawa iteration with DUNE-ISTL, and a solver-comparison workflow.
  • 3.3 Documenting an Interdisciplinary Workflow: The workflow documentation represents the solver study as processing steps that consume discretized linear systems and produce solved systems, runtime and memory measurements, and comparative runtime analysis.The first three steps share a reusable structure, allowing one documented solving step to serve as a template for others.
  • 3.4 Searching the MaRDI Knowledge Graph: Researchers can query all three data types through a guided search catalog that translates questionnaire criteria into SPARQL queries without requiring SPARQL knowledge.The resulting query and results are displayed directly in the RDMO interface.

4 The MaRDI RDMO Instance

The dedicated MaRDI RDMO instance offers a ready-to-use entry point for researchers, while the plugin can also be integrated into existing local RDMO installations. Its catalogs and support resources are intended to broaden participation in FAIR mathematical research data management.

  • 4 The MaRDI RDMO Instance: The MaRDMO Plugin and questionnaires integrate into existing RDMO instances, while the dedicated MaRDI instance lets researchers begin documenting without installing or managing infrastructure.Researchers can use either their familiar local RDM environment or the hosted service.
  • 4 The MaRDI RDMO Instance: The hosted instance is provided through DMP4NFDI, supports NFDI AAI login with institutional credentials or ORCID, and could be rolled out to other NFDI consortia.It is hosted at ULB Darmstadt as part of a multisite platform.
  • 4 The MaRDI RDMO Instance: The instance currently includes the mathematical-model documentation catalog and Basic Model Catalog, while algorithm, workflow, and search catalogs are available on GitHub for later deployment.The Basic Model Catalog provides a simplified entry point for model documentation.
  • 4 The MaRDI RDMO Instance: Support is provided through direct questionnaire feedback, a mailing list, monthly online meetings, and instructional YouTube videos.These resources cover the instance, RDMO, MaRDMO, and related data- and software-management topics.
  • 4 The MaRDI RDMO Instance: Together, the hosted service, integrations, catalogs, and community resources lower the barrier to documenting, discovering, and reusing mathematical models, algorithms, and interdisciplinary workflows.The stated scope includes researchers from all disciplines, including first-time contributors and users seeking existing resources.

5 Recent Developments

MaRDMO’s recent developments simplify model documentation and make knowledge-graph searches more precise. The Basic Model Catalog supports incremental, reuse-first documentation, while class-filtered search targets relevant existing entries.

  • Basic Model Catalog: The Basic Model Catalog lowers initial documentation effort by covering Research Problem, Mathematical Model, and Computational Task with lightweight formula placeholders.Researchers can later enrich entries through the Complete Model Catalog.
  • Basic Model Catalog: Formula placeholders require meaningful references such as a DOI and equation number instead of complete MathModDB-conformant quantities and quantity kinds.Academic disciplines, quantities, and quantity kinds are omitted from the basic catalog.
  • Class-filtered search: Class-filtered search restricts results to items carrying a specified instance-of statement, replacing broad autocomplete results from all item classes.This targets searches for mathematical models, research problems, algorithms, software, and other classes.
  • Class-filtered search: The search improvement gives researchers a more targeted view of existing knowledge-graph entries, directly supporting MaRDMO’s reuse-first approach.The stated benefit is improved precision and usability.

6 Outlook and Conclusion

MaRDMO is pursuing context-aware chatbot guidance and publication-to-questionnaire extraction to reduce manual documentation effort. The paper concludes that its RDMO integration makes FAIR documentation and discovery practical while supporting cumulative reuse across disciplines.

  • LLM-assisted developments: An import plugin is being developed to extract model, algorithm, or workflow information from publications into a reviewable version-zero MaRDMO questionnaire.Researchers can review, adjust, extend, or partially discard the generated documentation.
  • LLM-assisted developments: A context-aware RDMO chatbot is being integrated to provide role-specific guidance for DMPs, SMPs, mathematical models, algorithms, and related documentation tasks.The intended expert context changes with the researcher’s active documentation task.
  • Conclusion: MaRDMO connects researchers to the MaRDI ecosystem without requiring expertise in knowledge graphs, ontologies, or SPARQL.Embedding the functionality in the familiar RDMO tool is presented as making FAIR documentation realistic for everyday research practice.
  • Conclusion: The Stokes-Darcy solver comparison illustrates that one study can span mathematical models, algorithms, and interdisciplinary workflows within MaRDMO.The reuse-first approach lets researchers build on documented entries rather than starting from scratch.
  • Conclusion: As community contributions grow, documented models, algorithms, and workflows become more findable, reusable, and connected across disciplines.The paper frames FAIR in-silico research as a cumulative rather than one-time effort.
Loading 2609.11931v1…