[Data-modeling] Community (repost)

spencer kelly spencerkelly86 at gmail.com
Sat Feb 14 23:14:09 UTC 2009


It's a tricky situation then.  I understand Freebase is more
vulnerable than other crowd-sourcing projects and also has a higher
learning curve.  Makes sense.

However I think we all agree that it is not reasonable to expect every
creator of a type or base to stick around for it to mature.  A lot of
our types are floundering. It would be interesting to see a MQL query
of types whose creators haven't logged in for months.  We all agree
that forking or denormalizing (our current approach) is not ideal.  A
copy-button scheme (which I agree we need!) would absolutely trash
this project.  Imagine trying to make sense of freebase when there are
3 deceased person types or something.  I would be a devastating
failure, considering what we're trying to do (and what the internet
hopes we can do) is build some sort of ontology mountain.
Non-collaborative schema building will struggle to build something
this classy and comprehensive and this is our most pressing problem,
and one that will get worse.

so new proposal- Some sort of user graduation scheme.

I 've heard talk about giving top users gardening permissions. I now
suggest that users with certain community recognition should be able
to edit other people's types. Though I am no deft data person, I don't
see this as being overly dangerous.  I don't think adding a property
to a type would break the other properties, (would it? and if it did,
could this be remedied?)  and we'd of course have some sort of formal
consensus deletion process like wikipedia, and especially for changes
that would result in serious loss and mess things up. Really,
situations of this short would be rare i'm sure, and what a change
like this would mean really is that if I want to recip a property I
can. The role a type will play in the scheme of things (har) will be
subject to community consensus, not just the interests (and
attention-span) of it's particular creator.

For the particularly nervous and brooding, maybe we could have a
private option (like a draft type) that allows for total personal
control, but could be copied (only once!) with property delegation
into an admin-editable version if it has a niche.

A public/private distraction would also provide a way to deal with
'movies I like' types without discouraging potentially useful
contribution.

Once a user makes their type public it will be adjusted to reflect
agreements between knowledgeable people and the existing work that has
been done. Devastating edits (which are admittedly now possible), can
be minimized a number of ways (and reversed?) and through
collaboration we will naturally see a new pressure for consistency
among our work.  A decision for Freebase to give edit power to a
consensus of experts will allow for the long-term development of one
robust schema - goal I feel is threatened by the current process.
yes/no/good/bad/crazy? please toss in your many thoughts~
cheers-


More information about the Data-modeling mailing list