Raw

They're back on the agenda (as if they ever left), Paolo's had a look at some of the blog taxonomies Tim Bray points to.

I agree with the general consensus that there's a role for sharing personal/local/domain-specific taxonomies as well as more general (DMOZ-style) ones.

The main use for this kind of thing I'd say is narrowing down the search space for information - getting closer to explicit choosing: the category "Paolo's blogs" has two entries : English and Italian.

The problem is mapping between different taxonomies in a useful way, so we can e.g. "compare opinions about the same topics expressed by different authors".

I don't a simplistic 'big tree' approach can work here, because the structure as a whole is way too arbitrary for a tree. Take the example above - how would you link my /Paolo/blogs/English to a more global /blogs/English/Paolo?

It's true that you don't need fancy languages for useful tools in this space, as k-collector and Technorati demonstrate.

But (IMHO) to take real advantage of data already described in distributed, arbitrary taxonomies a fairly flexible description language is needed ( can you tell what it is yet, kids?).

Groves would probably do, but there seems to be little activity down that line, which probably leaves just Topic Maps and RDF. TMs could certainly take it, but my centissimi are on RDF, especially because folks are busy working on this kind of problem now. The ESW RDF Thesaurus work (Issue 9) should do nicely.

Once you can express this stuff in RDF, then off-the-shelf RDF tools (like Jena) can reason about it, and give you the kind of utility Paolo's talking about, for a lot more than just blogs. It's easier and can potentially go a lot farther.

PS. I just visited Norm Walsh's blog, he's doing some very tasty stuff with RDF there. I'd seen the very cool facetted filtering/ordering/taxonomies he'd got set up ( Dates | Coverage

Topics | Subjects) what I hadn't seen before is his handling of photos. A post about a pump with a photo together with cartloads of metadata. Much of this has been extracted from the RDF provided by the camera and stored in the JPEG, extracted using Norm's own jpegrdf. We are not worthy!

PPS. Semantic blogger Steve Cayzer has a related post - it seems Dave Winer* has invented ontologies.

Personally I doubt that OPML, with its single semantic of "child" is quite good enough.

* link removed

[Danny Ayers]