I've been noodling around with an idea but it will rely pretty heavily on Stardog's support of SameAs and I was wondering what impact SameAs has on performance. Should be pretty minimal right?
I haven't really worked too much with SameAs but it does seem to have its idiosyncracies. A fixed canonical URI is nice but it's not stable. Logically fine but UX is going to be terrible especially when URI's from different namespaces get mix and matched. It would be nice if there was a way on the client side to a rewrite and give weights to different SameAs namespaces. I guess you could do that completely outside of Stardog.
We have only ever recommended SameAs reasoning for small graphs, as the performance degrades as graph size increases. You will notice in the docs that it is not included in our next generation reasoner, Stride, as few people have used SameAs reasoning. As you point out, the experience isn't great as owl2 SameAs basically requires you to do all the hard work yourself, and then you have a special edge to say two nodes are the same. In practice, very few people have ever deployed SameAs reasoning.
What's more valuable is the hard work and seeing those weights or probabilities on two records being the same, and then being able to express rules to classify nodes in certain ways. That's the rationale behind the Entity Resolution feature, which produces that matching graph. You would then create rules around different matching values to further classify nodes, or create equivalent links, fuse the data, etc.
Thanks. Then I'll steer clear of SameAs as I take a look at an idea I had. I'll make a follow up post on what I had in mind.
I've been interested in poking around with the stride reasoner. There's something stupid that I've wanted to be able to do for ever that I think I can finally do with stride. It would be awesome if someone with a bigger brain that ,I like @pavel, would explain what is implied by having stratified negation.