Research Data Engineering Calling Mathematics

Himpe, Christian

doi:10.5281/zenodo.17831950

Research Data Engineering Calling Mathematics

Himpe, Christian

Publisher: Zenodo
DOI: 10.5281/zenodo.17831950
Repository: zenodo.org
Language: en

Abstract

Those not learning from past mistakes are doomed to repeat them – As a former research mathematician, I will summarize examples of past fails from mathematics in research data management and research software engineering. In my current position as a data engineer at the State and University Library of Münster, I will propose some simple measures for research data, software and publications from which not only mathematicians but all scientists could benefit. The insights leading to these propositions arose during the construction of a new metadata aggregator called "DatAasee", which will index metadata of research data as well as bibliographic metadata from distributed data sources. This is an open-source tool for libraries to provide one API to all their (meta-)data, and will hopefully help to avoid reproducing former tire fires.

Full text

Lines and areas can be changed in line with the Corporate Design rules (see Corporate Design manual) GAMM FA RSE & RDM Kickoff Meeting (TU BS) 2025-12-05 Research Data Engineering Calling Mathematics Christian Himpe (University and State Library Münster) 2 About Me and Why I Care ●I am an applied mathematician by training. ●I published and reviewed papers presenting computational results (a few times). ●I co-authored doi:bsb2 (RRR) and doi:g4n4 (SSS). ●Currently, I am data engineering with research data in a university library. ●My task is to make all the university’s research artifacts discoverable ... DatAasee – A Metadata-Lake for Libraries ●Metadata Aggregator / Union Catalog ●A virtual data-lake & metadata catalog ●Think: an organization-wide data search engine ●Repo: github.com/ulbmuenster/dataasee ●Paper: arxiv:2409.05512 OAI-PMH S3 DataCite DublinCore LIDO MARC MODS DatAasee 3 4 Old Data in New Bottles ●Research Data Management vs Research Software Engineering ●Code is data, so research software (source code) is research data. ●Data can be code, so it may be useful to treat it like code (sometimes). ●However, data and code should be available. ●Long standing problem: On Reporting Computational Experiments with Mathematical Software (1979) 5 Fail 0: Source Codes are not Treated Like Proofs ●Why is that? ●They are like a proof (Top Ten Reasons to Not Share Your Code) ●And without, … ●… how to replicate? (verification) ●… how to reproduce? (validation) ●… how to reuse? (application) 6 Fail 1: Legacy Benchmark Collection ●We preserved a research data set into a community Wiki (MORwiki). ●Why? Make data available again ✔ ●Original website gone, but it and the data was archived. ●What are the data-set licenses? Unclear ☹ ●What are the description licenses? Very unclear ☹ ●Best effort contacting the original authors. 7 Fail 2: Using RunMyCode ●I used RunMyCode for supplemental source codes. ●Why? Permanent link (but no DOI), Sloane funded (like arXiv), trusted advocate ✔ ●Let’s check a link: runmycode.org/companion/view/3760 (archived) ●Website gone … I guess the organization ceased ☹ ●Luckily, I have all the codes on my personal website. ●No notice to authors ahead of time! 8 Fail 3: Publishing my Thesis as a Book ●I published my PhD thesis with the Sierke Verlag. ●Why? Book, E-Book, DOI, Amazon, Google-Books ✔ ●Let’s check the DOI: doi:10.14626/9783868448818 (archived) ●No email answer, website broken … I guess the publisher ceased ☹ ●Fortunately, I negotiated secondary publication rights. ●No notice to authors ahead of time! 9 Fail 4: Using Papers-with-Code ●I briefly used Papers-with-Code. ●Why? arXiv-integrated, visibility by linking, ML/AI people use it ✔ ●Let’s read the news: Papers with Code is joining Facebook AI (via TIB) ●… ☹ ●Thankfully, I have all the codes on my personal website. ●At least some notice on time by the article above. 16 But Really ●Learn from past mistakes. ●Assume all platforms fail (Lindy Effect). ●Think about your work’s metadata. Your Questions Here … (Slides at himpe.science/talks/himpe25-gammfa.pdf)