Open Science Graphs interoperability: data models and data exchange protocols

An Open Science Graph (OSG) is an information space describing through metadata one or more entities and actors involved in the research lifecycle and knowledge production (projects, publications, data, software, services, researchers, organizations, facilities, etc.). 

The work of the Interest Group (IG) focuses on open challenges in Open Science Graphs for FAIR Data. As motivated in the case statement of the IG, the topic is of high interest and priority globally, with this specific RDA IG featuring in several European Commission calls for projects in 2022.

During the last IG session at RDA, it became clear that the most urgent challenge targets the definition of an interoperability framework enabling a seamless exchange of data across diverse OSGs. According to the audience, whose composition spans 40% researchers, 30% service providers, 30% graph providers, developers, and others, the following high-priority aspects were identified:

• Identification of data sources (e.g., registries, Open Access archives, as well as other OSGs) integrated by the OSG at hand;

• Implementation towards Openness and FAIRness;

• Common export format and protocols: identification of a set of research entities, inclusive of publications, datasets, software, organizations, funders, authors (as well as services, patents, research concepts/topics, and annotations, here less relevant, but yet of interest);

• Specification of added-value specific to the OSG at hand.

The objective of the session has the main objective of bootstrapping the discussion about the establishment of a specific Working Group (WG) discussing on “data models and exchange protocols for Open Science Graphs interoperability” that would tackle such challenges. In particular, we foresee a discussion centred around:

• Data model

  • How to identify of objects in the graph: graph IDs and PIDs together with a vocabulary of PID providers (Crossref DOI, DataCite DOI, ORCID, ISNI, ROR, ArXiv, PubMedID, PDBs, etc.);

  • Types of entities: vocabulary of entity types;

  • Entity description: minimal (and optional) set of properties for each entity;

  • Relationships: scholix.org;

  • Provenance of the object, provenance of the object properties;

  • Trust of the object, trust of the object properties; trust should be a value (e.g., from 0 to 1) indicating the quality/accuracy of the object or property (e.g., 0 = randomly generated data; 1 = data manually verified by an expert agent in the domain);

  • Added-value: set of entity properties characterizing the added-value brought by the Graph (e.g., indicators, subjects, references, etc.).

• FAIRness and Openness Definition of a “graph profile”, an exchangeable metadata description of an OSG describing in a machine-readable format all the properties above, enriched with:

  • Integrated data sources: refer to a registry of data source/service PIDs;

  • Licensing information;

  • Documentation: metadata schemas of the entities added-value, access protocols, list of dumps (and related PIDs).

The first part of the session will be dedicated to a discussion on this approach and of possible alternatives, with the intention of identifying broad agreements and disagreements on the WG topics to build a solid starting point to submit a request for an RDA WG. 

The second part of the session will focus on the definition of a roadmap for the WG, identifying a list of properties, contributors, and meetings during 2022. An optimal scenario is one where the group marries an incremental approach, defining a core Interoperability Framework that can be quickly adopted by OSG providers as a starting point to facilitate and track data exchange, followed up by richer capabilities, to be adopted at a later stage to further enhance the resulting ecosystem.