MARDI4NFDI

The Mathematical Research Data Initiative (MaRDI) uses Wikibase to construct a comprehensive knowledge graph that interlinks diverse mathematical research data. Accessible via the MaRDI Portal, this knowledge graph integrates over 7 million items and 700 million relationships from various open sources. This approach supports the FAIR principles by assigning persistent identifiers to each linked resource, enhancing the findability and interoperability of mathematical research data. Researchers can utilize the MaRDI Knowledge Graph to discover related datasets, software, or research papers on specific mathematical topics, and explore links between mathematical models, equations, and computational methods through structured queries via an API or the MaRDI KG Query Service.

We are using a Wikibase instance at https://portal.mardi4nfdi.de/ to aggregate, integrate and curate research-related data across the field of mathematics. We chose this model because it allows us to

• customize our workflows and infrastructure for mathematical needs, while staying closely aligned with Wikidata
• handling data that would not fit the scope of Wikidata (e.g. in terms of granularity).

Besides some basic on-wiki documentation, we keep our Wikibase-related code on GitHub at https://github.com/MaRDI4NFDI (particularly in the https://github.com/MaRDI4NFDI/portal-compose repo), and we wrote up an overview of our progress so far at https://arxiv.org/abs/2309.11484. 

Our project is ongoing, and we expect that we can adjust it to our most important needs. Some of the challenges that we encountered and that may be relevant to other Wikibase instances:

• While most of the individual MediaWiki extensions, MediaWiki templates, Lua modules, gadgets, bot scripts, policy pages, namespaces etc. from Wikidata (or any other Wikibase, for that matter) can in principle be adapted in a rather straightforward fashion, the ecosystem as a whole that is composed of all these parts – or even just the major ones – is hard to adjust or even to recreate.
• Some of the functionality that we need does not exist in Wikidata or any Wikibase. For instance, we want to be able to cite the results that a specific SPARQL query to our endpoint returned at a specific point in time.