Guillaume Anciaux (EPFL)

Academic knowledge is traditionally disseminated by academic journals. However, nowadays, the production of scientific data in any given project exceeds by a vast amount what can be contained in a few journal pages. Reproducible scientific data and publications must be associated to boost scientific collaborations and discoveries. There is today an always increasing pressure coming from universities and funding institutions towards publishing open data. Such a curation may represent an important workload, floating in between librarian tasks while always needing the expert criticism of the specialized scientists. Despite these points, the Computational Solid Mechanics (CSM) community (and others), have only limited solutions/tools to curate and share data produced by experiments and simulations. Furthermore, the current open publication procedures of private editors only too rarely involve curation beyond the paper article, and this despite the expensive Article Processing Charges (APC). At large, this situation does not favor open science.

In this presentation, I will start with a small history of the academic press, up to the diamond open access principles and the role that can be played by scientists themselves. Naturally the evolution of practices towards digitization and computational disciplines will bring us to consider datasets.  The possibilities and the rather limited curative requirements for datasets in generalist repositories such as Zenodo will be detailed.

I also will present the recent initiative of the overlay diamond open access "Journal of Theoretical, Computational and Applied Mechanics" towards the curation of the datasets accompanying published paper. In the same spirit as for the paper review procedure, it is believed that datasets can only become valuable when correctly cleaned, annotated, documented and proved with minimal reproducibility. This move was made possible by the joint developments of the DCSM project, which provides a web platform deploying an open-source software (Solidipes) to support computational solid mechanics (CSM) scientists in curating and publishing their datasets. Such a platform will be briefly presented and illustrated with show case papers+datasets.

Finally, I will depict what should be the ideal curation framework, to be deployed in institutional repositories as well as in the back-office of diamond open access web platforms (e.g. Episciences). In particular, the need for specific ontologies, describing the relational hierarchy between objects, will be linked with robust and simplified recognition/validation procedures. In other terms, we call for a normalization of the (discipline dependent) scientific community output. While this certainly represents an important amount of work, it has the potential of reducing the workload for dataset curators. Which is a sine qua non condition to convince dataset reviewers leveraging the value of our digital production.