Re: Tag ontology

Seth,

> It would be nice to have this ontology.  One thing i would like to see
> in it would be something akin to a WordNet synset so that we could
> assert that some tag is a member of a set or synonyms.   The triples
> might look something like this with the topic being a URI to the
> preferred tag according to somedomain.

Synonymity, of course, is quite subjective (not to mention 
context-dependent), so subjectivity would have to be borne in mind --- 
tagging something with "ruby" should not automatically throw it into 
synonymity with "rubylang" thanks to some outside synset. Working 
around that is a challenge, and I would suggest the approach of using 
strictly local tags (by which I mean "in my own namespace, not someone 
else's"), using URIs rather than literals, and then mapping them to 
others' when identity is established.

A very practical way would be to use del.icio.us directly --- after 
all, my use of the term "objectivec" is encapsulated in the URI 
<http://del.icio.us/holygoat/objectivec>, which is under my control.

RDF hangs nicely off that, so as you say

> <http:://somedomain/?synset=folksonomy> contains "folksonomy".
> <http:://somedomain/?synset=folksonomy> contains "folksonomies".

I can relate <http://del.icio.us/holygoat/objectivec> to 
<http://del.icio.us/holygoat/objc>, whether through a direct 
relationship or some indirection (like a synset). I think there are all 
sorts of relationships here, from synonymity and other clear meanings 
all the way through to the basic "related tag". All the way you'd be 
butting your head against the ambiguity of tags, of course, but hey --- 
that's folksonomy for ya! :D

> If we had such a database we could use it right now.  What would be
> even keweler would be if delicious et al would allow us to create that
> db in the process of tagging.   You should be able to grab somebodies
> synset and process it *interactively* against your own.   Am i getting
> ahead of myself here?

Possibly! :) It's not _too_ far ahead, though; once we have a tag 
vocab, and a source for tags (be it del.icio.us or something else), 
then hanging smarter info off them is straightforward. I don't think 
del.icio.us offers quite the facilities that many would want. A 
tag-based browser that could use these richer relationships would be a 
step towards a more usable del.icio.us.

Here's an idea that just jumped into my head: how about a "synset" tag 
on del.icio.us that tagged RDF files containing everyone's synsets... 
that would be a data goldmine!

I shall get cracking.

Thanks for the great feedback.

-R

Received on Tuesday, 22 March 2005 18:28:25 UTC