@rowland.watkins Thanks, I actually didn’t think of that. However, I wonder what kind of complexities I could run into with named graphs since I also make heavy use of inferencing in Jena. Maybe it won’t be an issue at all. I guess this is the best way to prepare data for export. In any case, I still want the metadata relations in place since the dataset and all relevant schema are meant to be browsed:

While my knowledge of the tooling is definitely behind (I haven’t used Jena in over a decade), I used the NG4J (Named Graphs 4 Jena) extension lib to do a lot of the heavy work. If memory serves, you can specify the Model Jena outputs entailments to, in this case a NamesGraphDataSet. Jena has clearly grown over the years!

Throwing entailments into a names graph was very useful to me since I was able to garbage collect the forward reasoner output between runs

@simongray so in your case there could be an option to throw entailments into a named graph and optionally merge them with the main dataset if they are useful