Hi Perter,

> I think it is reasonable to tag anything with anything myself. I don't
Totally agree.
> think tagging a document on the internet needs to be associated with a
> skos concept. I think you should be able to derive the tag from
We are talking about the "tag", not the tagged thing here (think it is 
where we are getting lost)
> anything, although if it had an rdfs:label and a rdf:type it might
> help, but ideally you just need to have the link to the resource and
> know that it is a link. Unless you want tagging to become so
> cumbersome noone uses it I think there should be a focus on simplicity
> and interoperability with the ability to tag at the most basic level
> of resources to resources.

We are always talking about MOAT here, with the same architectures, the 
same systems. Don't forget that MOAT is supported by a server side 
system that helps users to relate a tag to meaning(s). It is a 
semi-supervised system.

We are, in fact, talking about how to describe what Alex put in his 
system: what is the best way to describe the things he envisioned (the 
meanings, the tags, etc)

> If you leave it at the generic Resource level than tagging should only
> contain one extra step to normal, ie, choosing the URI from any that
> they know about. If you enforce a narrow range then they are quire
> likely to have to create a new resource before tagging can continue,
Depends on the system.

It is really about the semantic of the vocabulary: what is a meaning? 
How to relate a meaning? how to describe a meaning? etc.
> or they could just tag without semantics. Personally I haven't noticed
> many ontologies that are encoded with the SKOS scheme, so enforcing
> that would seem excessively restrictive also.
Taxonomies, not ontologies (SKOS is a taxonomy that helps people 
describing taxonomies).

Excessive I am not sure; not the best thing to do: possibly. We are 
really discussing about pros and cons here.

> How do you envisage the user process for this to occur? Does someone
Exactly the same way as the current MOAT.
> have to build a category/subject for every tag they may want to use?
Same as here. A dataset is used for this (currently DBPedia)
> Folksonomies are successful because they are easy and generic. Normal
> people like simple.
Yeah, it is as simple as the current MOAT.

> I might be confused by the references to concept. As I see it you are
What I am thinking too :)
> giving meaning to a text-string, which may not be an abstract concept
> at all. It could be a concrete thing. Which seems to be at a higher
> level than skos:concept, which seems to only include abstract
> categories. I don't see why owl:Thing can't replace skos:concept, and
> why from there you can't go to replacing it with a generic
> rdf:Resource range. Folksonomy tag's aren't only abstract is what I am
> trying to get at.

Question: can a named entity mean something (so, being related as the 
essence of a meaning (like in MOAT)).

I have some doubts about that.

So, does it make sense to say: I have a tag "physics" that has a meaning 
that is represented by the URI of "Albert Einstein"? Does Albert really 
*mean* "physics"? I don't think so, but this could be possible with the 
current ontology.

This is what I am thinking about right now.

Take care,


> Peter Ansell

Received on Monday, 21 January 2008 01:36:25 UTC