From: Jeff Bone (firstname.lastname@example.org)
Date: Mon Dec 31 2001 - 12:34:36 MST
Ben Goertzel wrote:
> It can be done to some extent, see for instance the work of rzhetsky at
This is interesting stuff... wasn't aware of it. Thx!
> How are you going to
> get the original datasets out of the research divisions of pharma companies??
Hmm, well, that's not my experience --- in either academia or the private
sector. (I recently got some really interesting filesystem usage data from a
Microsoft Research effort, for example of the latter.) But that's likely an
industry / culture thing,
> About this aspect of the sociology of the research community, I'm afraid you
> really don't have a clue.
Easy there, hoss --- different experiences lead reasonably to different
conclusions. And btw, while we're being naive: let's assume that you *can*
extract such useful information from summaries. Why do you believe that this
won't just cause pharma, etc. to clam up and not reveal *any* information?
I.e., even if the technology is feasible, to the extent that it's available it
will just trigger additional competitive concerns and restrict open publication
(There's a completely different discussion we can have about the economics of
"soft property" and whether any sustainable business model can be built on
secrecy. But another time...)
> I still believe that information extraction from
> NL text will be an important part of the toolkit too.
No doubt. My concern isn't that this isn't important, it's that having it as a
sole --- or even primary, ahead of the other things --- focus is less useful at
best, dangerous at worst.
> Check out
Yeah, yeah. We'll see how widely used it becomes. (I've become rather cynical
about standardization and commercial interests since some rather bad IETF
interactions re: event notification / instant messaging a few years back.
Notice that MS Messenger, ICQ, and AIM aren't all fully / continuously
interoperable, even now? Notice how the lock-in helps the vendors on a
competitive basis? Same thing for any standard, unf. And for things --- like
online communities, and like information about genes and protein expression ---
where there are "network value chains" this effect is even more pronounced.)
This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:00:37 MDT