RocketRML is an RML mapper implementation that efficiently maps data from various sources to RDF. In our use cases for e-tourism, the IT solution providers serve their data in XML or JSON format. The standard RML implementation requires joins, which needs Primary - Foreign Key pairs in the data objects to be joined. 

The example above however has a situation where two related data objects do not have such a key pair. The only heuristic that indicates the relationship between a ski resort and its addresses is the nested structure. In order to overcome this issue, RocketRML implements a custom iterator that does not use joins but it recursively iterates over nested objects and maps them one by one instead of mapping all resources and eliminating the wrong relationships, as it is done with joins.

The RocketRML performs significantly better in comparison to standard RML implementations. This is mainly due to the fact the RocketRML does not have the overhead of joins. The results of the comparison can be found online. The full paper can be also read on the workshop website.

What are the indications of this work for semantically annotated Web APIs? First of all, such a mapping tool can be used to convert a Web API description in a format like JSON or YAML (e.g. Swagger) to a schema.org actions annotated API description. We are investigated further usage at the moment. 

Hopefully, we will have a nice and productive workshop in Portoroz on 2.6.2019 an discuss the future directions in building knowledge graphs.