Revision as of 01:39, 1 November 2007 by Admin (Talk | contribs)
(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)


How should we proceed?

Most recent 'root comment' first (conversation threads in descending order):


How about starting by enumerating the properties of the most popular existing sites, then seeing what kind of class hierarchy "falls out"? This gets us moving (embrace and extend!), and maximizes usability, as people, when migrating their profiles, can just choose the "view" with which they are most comfortable (FOAFster, FOAFLinkedIn, FOAFTribe, etc. -- all of which could be sub-classes of "Core FOAFster"). -- Frank Ruscica


For the moment I'll copy over my e-mail to rdfweb-dev; when I get some time, I can break things down a bit. -- Nick Knouf

Just for starters, let's think about the properties that you can enter on a site like Friendster (and it'll be no secret that I'm talking about my own profile, but that's not pertinent to the thread :-) ).

Friendster offers free-text input of favorite music, movies, books, and TV shows. Separate items are delineated with commas and form clickable search terms to find others with the same interest. This seems to be a perfect place to include some of the ideas of the Charette Relationship Set (CRS) [1], which can describe things that one has or wants, hobbies, and stuff that you're interested in. There should be some tie-in with other media-specific schemas (and I'm not saying this is what should be used, but stuff like the MusicBrainz Metadata Initiative [2] for music could be used).

Okay, there's a whole range of work to be done with the media-specific schemas, but let's move on to more superficial things :-)

Physical attributes like hair color, complexion, and the like probably best come from governmental work like Stephen is looking at. I'm not sure of previous efforts to formalize definitions of someone's height, weight, etc.; with those properties, however, we need a good way to describe the units. Someone looking for information on me would want to know if my weight of "150" was in pounds or kilos ;-) I asked about this on www-rdf-interest in December [3] and was referred to some work done at JPL [4] to describe the units in a very elegant way. Making this work in practice and making the ontology validate as DL is somewhat more difficult, but I'm sure it can be worked out.

Geographic information can probably come from existing ontologies (which others on this list have much more experience than me).

What doesn't seem to be out there is good semantic definitions of "Interested in meeting people for" (e.g., dating, serious relationship, activity partners, etc.), "Who I want to meet", and so on. That's probably the crux of the problem. This is going to run into all sorts of cultural, religious, and geographic concerns as well; not that it means we shouldn't try it, but it's something to keep in mind.

I do believe that this is a case where pulling terms from a large number of ontologies and namespaces is going to be useful and perhaps necessary. Describing a dating service is basically an attempt to fully describe a human being, much beyond what FOAF does (and I think because of that most of these additions or samplings should not be rolled into the FOAF namespace). I'm afraid that using a single namespace is going to make this blossom into something like SUMO [5]; I think SUMO is very interesting, it's just not what we're looking for (in my view).

We ought to be realistic here: after thousands of years of the written word, no-one has come up with the _ideal_ way of describing someone. Everything we do will only be an approximation. Of course, we should try and make it a damn good one :-)



[1] [1] [2] [2] [3] [3] [4] [4] [5] [5]