During the first year of the project, SPRINT partners analysed the landscape of the IF provided by different Shift2Rail projects. The gathered requirements, together with those provided by the analysis of similar initiatives in different domains, have driven the design of the first version of the SPRINT IF architecture and helped identify the priorities for the development of the technical elements of the architecture itself.
The architecture defined the IF as a set of interconnected tools and services, all of them based on the assumptions of using Semantic Web technologies:
• the Converter architecture, which has been designed around the concept of “conversion pipeline” which is commonly used in ETL systems. Flexibility has been taken into account by providing the possibility to use different strategies for converting data into RDF and to extract structured data from RDF data.
• the Asset management solution, which features state-of-the-art governance management and can enforce complex processes orchestrating human interactions and services. By integrating both a BPMN solution and a continuous integration and deployment (CI/CD), our lifecycle processes support a wide array of automation possibilities (i.e. generating documentation for ontologies or converting datasets into different formats / specifications exploiting converters). We studied also the possibility to use this component as a “configuration server” for Converters, which were able to dynamically obtain from the Asset Manager mappings and other resources required to perform message conversions between different standards/specifications;
• the Distributed SPARQL Endpoint, which provides a uniform querying interface for distributed data leveraging on data virtualisation and query rewriting.
• the Collaborative ontology engineering tool, which enables the usage of shared code repositories as the basis for collaborative editing of ontologies. To ease the understanding of complex models, the tool is able to re-use state of the art automatic ontology representation and documentation tools to provide editors with up-to-date diagrams and human-readable ontology descriptions.
• the Mapping suggestion tool, whose aim is to shorten the time required to create mappings between data models and ontologies. It does so by applying machine learning techniques to suggest possible mappings, letting users create complex ones, with the ability to generate Java annotations as defined by the SPRINT project as the mapping outcomes.
The development process undertook two iterations of design, implementation and testing. Testing was performed considering both the functional requirements implied by a set of real-world scenarios defined in the project in accordance with the Connective project, and performance and scalability requirements.
