tagtriples

Hi all,

I'd like to announce 'tagtriples' - a experimental format for
exchanging and aggregating structured metadata. It's based on RDF
ideas, but emphasises simplicity over precision.

I've built it because I found that people at work struggled with
publishing their data using RDF, and I wanted to see if giving them
something simpler would yield more published structured data.

In particular, it has these differences to RDF:
- No URIs  (instead it uses 'tags' in the statements)
- No Literals (everything is a tag)
- Graphs are central to the model 
- No BNodes*
- No collection types*

I'd like to emphasise 'experimental' - this hasn't had much testing
and although I have some ideas about resolving tag ambiguity issues,
these haven't been implemented or proven yet.

Sparse web page:
http://tagtriples.sf.net/

Thoughts about the format on my blog:
http://www.phildawes.net/blog/
http://www.phildawes.net/blog/category/semantic-web/tagtriples/

I've written some tagtriples aggregator software (see sf page) with
web based browsing and a sparql-like query language. 
I've recently deployed this at work, it scales to >1M triples with
millisecond queries (should scale much higher, but not tested) and
also supports very fast substring text querying using a suffix array.

There's a (slow) demo of this software on my website:
http://phildawes.net/temporary/tagtriples/tag/FrenchHorn

(N.B. There's problems with the python binary crashing on the hosted
server - if you get a 500 just refresh)

Hope this is of interest to somebody!

Cheers,

Phil

* these decisions aren't set in stone

Received on Wednesday, 23 March 2005 13:54:09 UTC