[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [kDev] Why KendraBase?... (was: The idea could be good...)



On Tue Nov 18 16:23:47 2003, Daniel Harris wrote:
On Nov 14, 2003, at 8:44 pm, Dave Cridland [Home] wrote:
Daniel, how does what's being built *now* - a semantic cloud of free
and easy data - relate to a collaborative design system? ... I'm not
even certain exactly what a collaborative design tool exactly is

Don't get too hung up on the grandiose terminology. Together with what
I've said above read specifics in the project plan here:
http://www.kendra.org.uk/wiki/wiki.pl?KendraToolsProjectPlan

Has nothing to do with KendraBase - this document actually includes KendraBase in its requirements, which I'm not accepting as a reason to use it - "Well, it's here in the requirements, so we *must* need it!".

Otherwise, there's little that couldn't be done considerably better by a specialised system - lower applicability, but higher optimization and easier development.

And specifically the "live demo" here:
http://www.kendra.org.uk/wiki/wiki.pl?KendraBaseScenarios
and let me know if it doesn't answer your question.

1) Addressbook.

Solved already with existing prior art (ACAP + LDAP). In this case, your addressbook is also held on a server.

It's not a widely deployed system, but the system does exist.

2) Fuzzy fluffy search stuff.

Sure this is interesting, and sure it's applicable to Kenrda's goals. But what I *don't* see is how it's more applicable than a more specialized system, which would be easier to build and deploy, and is likely to be considerably faster.

3) A relevant example.

Note that every example you've given is fixed language. The artist "Pink Floyd", track "Another Brick In The Wall Part 3", Album "The Wall", is still in English even in France. The reviews may well be in a different language, but I really fail to see how you'll solve that one.

Genre is the interesting exception - you could translate these automatically by seeking correlations, but you're entirely likely to find that the genres don't correlate between catalogues in the same language, let alone between different languages. Assigning this to the end-user lowers the correlation rate, but increases the chances of getting it right most of the time.

Trusting someone else's judgement on what a genre is can be done by trust networks, which can also be used for rating assessment, etc - but we've discussed all this before.

(Oh, and you can of course, do some weird stuff, like occasionally asking users to pick a genre for a track, say, and from that deduce what genres they'll pick for other tracks - or even what music they'll like, or what books they'd enjoy reading.)

Again, this is something that can be done by a specialized system far more efficiently than by something highly generic.

Oh, and I think you're missing a goal from the first paragraph - "find
a way to make people *want* to use the system". Without that goal,
Kendra is a worthless system.

I think we've already found the "way to make people *want* to use the
system". And it is to make the system useful to them. And so we need to
get their requirements. And so we must enable organisations and
individuals from every link in the content value chain to participate
in the creation of Kendra System. And because every participant gets to
put their requirements into the pot they get a system that will work in
the way that serves their needs.

And for those views, you'll need KendraBase? Surely the Wiki and the mailing lists should be perfectly adequate.

Agreed, you'd get more feedback from the general public - the key in all of this, really - if there was a system online that people could try out.

But in any case, the goal has not yet been met.

On Nov 14, 2003, at 4:39 pm, Phillip Temple wrote:
... From what I've seen, this has evaporated into airy discussions
about "connecting people" and "semantics" which is already being
covered by far more competant people in academia.

First, I admit that "semantics" is a general word for a lot of stuff
and so will tend to confuse the situation. As far as I can tell it was
first used on Kendra mid this year and only appears once in the project
plan. When I use the word I'm talking about meaningfulness. I'm not

"meaning", I hope. "meaningfulness" is something different.

talking about artificial intelligence, no, no. I'm talking about a
shared understanding that when I say "song" I mean the same thing as
you. Or when I say "chanson" (a French word) I also mean "song". Very
simple stuff.

But:

a) It's not simple in the slightest. The semantics associated with a word in English are rarely precise matches for any word in another language.

b) It's not needed anyway. (With the possible exception of genre matching, but I doubt cultural differences will make that worthwhile).

c) The amount of time spent working on a system to handle this could almost certainly achieve a rough-cut version of the Kendra System, which would gain significantly more interest.

Now, about these "competant people in academia". They are both
"competant" and they are "in academia". But that doesn't mean they are
going to provide solutions that satisfy Kendra's goals. We are using,
as I'm sure many of us do, ideas from academia. But we need to
*implemement* them into a system that works for us.

*Do* we need to?

Dave