HN Explorer

Who is using Semantic Web technology in production?

7 points by edraferi ↗ HN
Semantic Web movement has produced several technologies (JSON-LD, SPARQL, OWL, etc) that seem useful for data integration, but they don't appear to have much momentum. I'm struggling to determine if the Semantic Web is a justly abandoned bad idea or a quietly useful mature technology.

The main place I see Semantic Web technology in use is SEO [1] and Email Markup [2]. Is anyone using these technologies for more general data integration tasks?

[1] https://www.bing.com/webmaster/help/markup-validator-e9b66817

[2] https://developers.google.com/gmail/markup/getting-started

5 comments

[ 3.2 ms ] story [ 27.0 ms ] thread
PaulHoule
Here is my take on what RDF is

http://ontology2.com/the-book/rdf-a-new-slant.html

JSON-LD and SPARQL are fine, OWL as an inference system has some good ideas (implementable by macros on top of production rules) but it is not adequate to do what people need for data integration (multiply, divide, add, subtract for one thing.)

edraferi
Thanks, that's a solid resource that I wasn't aware of. It's interesting how you're generating Java stubs using your K Schema format.

I like the idea of using RDF as a universal data model. I buy the power of graphs in general, and RDF seems much more useful than a generic object-property graph thanks to built in support for namespaces and reasoners.

My hesitation about the concept comes from concerns around performance and implementation difficulty. It looks like you've built some solid tools to ease the initial modelling challenge, how are you addressing the performance challenges? That is, once you pull several data sources into a huge graph, how do you keep query speed up?

PaulHoule
The "huge" graph is only virtual. It exists in principal but you are not going to materialize the whole thing when it gets huge. Or if you are you are going to do it with stupid Hadoop methods, not by trying to grow traditional databases up.

Trying to get the whole world inside one "envelope" is scalability pornography, i.e. SAP HANA.

Realistically you create the indexes and data structures necessary for your workload, it is not that different from any other technology except in the case of RDF the math is known.

usgroup
I tend to think of it as a serialising format with nice properties rather than a base data model. Tool support sucks for most languages right now and there is very few users who'd know what to do with a SPARQL endpoint , further there is no out of the box UI or UI toolkit in the open source space.

DBopedia and other serious projects offer endpoints. And the nature of the format means one can federate queries across them . DBopedia is currently the defacto ontology/ starting point.