Hey Afredas,
> Despite that the examples you give can be replicated with rdf and
> spraql (forget owl).
I did this to show that the same RDFS/OWL rules can be easily expressed in
Gremlin. In other words, its a relatively trivial exercise to turn these into
Gremlin expressions:
http://www.agfa.com/w3c/euler/rdfs-rules.n3
http://www.agfa.com/w3c/euler/owl-rules.n3
(the hard part would be the optimization and organization of rules for speed,
but just the rules, easy).
> One might argue that OWL is a good example of
> over-engineering, as in practice only the owl:sameAs is used :-)
Ha. Yea---Linked Data is testament that OWL is too much and unscalable being a
reasoning system for "The Web."
> But I think the real power of gremlin is within the grateful dead and
> spreading activation example. This is a radically different way that
> can be also used for reasoning and inference.
Yes. Description logic reasoners are global in that that work over the whole
graph. Spreading activation models of inference are usually with respects to
root vertices -- thus, local. Most all the algorithms I work with are based on
some form of spreading activation.
You may be interested in the work of Pei Wang and evidential logic. Evidential
logic is scalable, local, and based on "human logic," not axiomatic,
mathematical logic:
http://books.google.com/books?id=vEtzi3-d4PAC
http://www.cogsci.indiana.edu/farg/peiwang/papers.html
> I mean that example really has to get more publicity as its something
> completely different and useful :-)
Ha. Yes. I always talk about it in my slides---priors/local rank
algorithms---but people still say: "Ooo graph database! Now I can do PageRank
and shortest path!" :)
I'm glad I found someone who likes the same things I like. Note: I have another
blog post coming up on graph-based recommendation where I will be using a
spreading activation variant to explain "the product/interest graph" as a
neural structure. Will be fun.
Thanks for your positive thoughts,
Marko.
http://markorodriguez.com
_______________________________________________
Neo4j mailing list
[email protected]
https://lists.neo4j.org/mailman/listinfo/user
