Rewriting the past: A multi-level approach to improve quality into NAKALA repository

Huma-Num is a French national infrastructure dedicated to SSH (Social Science and Humanities) research projects. In order to meet one of our primary missions, that is, to provide solutions to the SSH community to preserve research data, Huma-Num has developed over the last 10 years a repository named NAKALA. The goal at the time was primarily to secure research data. In that sense, it is a resounding success, with around 2 million files grouped in 800 000 deposits, and NAKALA has found its place in the national ecosystem. However, the quality of the data and metadata of existing deposits is far from perfect which is detrimental to the visibility and hence potential reusability of these datasets: but how to tackle this broad objective of improving data and metadata quality? Given the large amount of data in NAKALA, it is clearly impossible to process each deposit individually, and therefore, it was decided to adopt a multi-level approach. During the “Core Trust Seal” certification process, HumaNum was compelled to review a number of features associated with NAKALA. Documentation was completely revised to align with best practices and help guide users to submit better quality data and metadata, with a particular focus placed on the aspects of data preparation before depositing. To help users implement these recommendations during the process of data deposit, compliance checks were added, and autocompletion has been added to several metadata fields to encourage the use of controlled vocabulary. Three main levels of data curation were identified: 1) data whose quality is unreviewed by humans 2) data that are checked from a documentary point of view by humans 3) data already at level 2 that are also verified from a technical and archival point of view to be preserved over the long term using the platform of our partner CINES. For levels 2 and 3, in order to check the data sets, a network of data stewards and experts was identified across France and a “moderation” workflow was created. Researchers can now reach out to local specialists to help them improve quality and receive a quality label.The various actions described above are designed to improve the quality of future deposits of data, but what to do with the previous deposits? A study has been launched to examine the overall quality using multiple types of criteria: from the content of the title and the use of URIs in appropriate metadata to more complex cross-metadata queries. The main idea is to be able to determine indicators and thus build dashboards to have a continuous vision of the global content of NAKALA and also more specifically for users in order to encourage them to improve quality. The poster will review the results obtained by implementing these different approaches to “rewrite” the past and how these decisions impact the future developments of NAKALA repository and the evolution of the support to users.