RE: [cmunell] Digest for cmunell@googlegroups.com - 2 updates in 1 topic
Barry Robson
Seems like a syllogism, though not exactly, but it seems that it would for example, imply the use of a higher order “rule” in the Semantic Web, a “metastatements” in Q-UEL, and an aspect of program as opposed to data in PROLOG, and so on. Perhaps what is needed here is to see how NELL stands on, or handles, such matters.
From: cmu...@googlegroups.com [mailto:cmu...@googlegroups.com]
Sent: Thursday, September 15, 2016 12:06 AM
To: Digest recipients
Subject: [cmunell] Digest for cmu...@googlegroups.com - 2 updates in 1 topic
· IDs indicates an entity? - 2 Updates
parag...@gmail.com: Sep 14 07:06AM -0700 |
Bryan Kisiel <bki...@cs.cmu.edu>: Sep 14 12:08PM -0400 |
You received this digest because you're subscribed to updates for this group. You can change your settings on the group membership page. |
Bryan Kisiel
obvious to get a prototype system up and runninng to see if we could get
something like NELL to work. In this case, arguably the principle
motivation was to use mutual exclusitivity between categories as an
additional constraint on the learning problem on the notion that coupling
different "modes" of constraints would solve what was expected to be an
otherwise underconstrained learning problem -- if more different "views"
of a possible representation of the world agree on what is true, then
that's a more reliable way to keep the learning on track.
In fact I don't think we really know for sure how important the
constraints imposed by using a hierarchical ontology are. NELL's current
ontology and use of it is so flawed that I'm tempted to guess that it's
not critically important afterall. However, I'm pretty sure we have yet
to see as good results come from the times when we've used ontologies that
are not very hierarchical. It's just not something anybody has
specifically analyzed, I suppose largely because what we do now is "good
enough" and so not terribly high on the list of interesting things to
revisit.
There's been some work on replacing many of NELL's unsophisticated hard
constraints with something more akin to jointly modelling many of them to
resolve conflicts among possible truths. But nothing that has yet made it
back into the system. I would expect improvements to come from that
direction.
bki...@cs.cmu.edu
On Thu, 15 Sep 2016, 'Barry Robson' via NELL: Never-Ending Language Learner wrote:
> Seems like a syllogism, though not exactly, but it seems that it would for
> example, imply the use of a higher order “rule” in the Semantic Web, a
> “metastatements” in Q-UEL, and an aspect of program as opposed to data in
> PROLOG, and so on. Perhaps what is needed here is to see how NELL stands
> on, or handles, such matters.
>
>
>
> *Sent:* Thursday, September 15, 2016 12:06 AM
> *To:* Digest recipients
> *Subject:* [cmunell] Digest for cmu...@googlegroups.com - 2 updates in 1
> topic
>
>
>
> cmu...@googlegroups.com
> <%20%20https:/groups.google.com/forum/?utm_source=digest&utm_medium=email#!forum/cmunell/topics>
>
> Google Groups
> <https://groups.google.com/forum/?utm_source=digest&utm_medium=email/#!overview>
>
> <https://groups.google.com/forum/?utm_source=digest&utm_medium=email/#!overview>
> Topic digest
> View all topics
>
> · IDs indicates an entity? <#group_thread_0> - 2 Updates
>
> IDs indicates an entity?
> <http://groups.google.com/group/cmunell/t/9f6a5d6e0cfbac3a?utm_source=digest&utm_medium=email>
> parag...@gmail.com: Sep 14 07:06AM -0700
>
> Emphasising on the last point you made: "male implies person", Does NELL
> have a mechanism to internally create a new representation
> "concept:person:male:A" from the two concepts "concept:person:A" and
> "concept:male:A" in a scenario where either of these two representations
> were existing in the KB and the other one was found at a subsequent point
> in time.
>
> I am trying to know about the self-improving mechanism of NELL.
>
>
> On Tuesday, June 21, 2016 at 2:18:33 PM UTC-4, bkisiel wrote:
>
> Bryan Kisiel <bki...@cs.cmu.edu>: Sep 14 12:08PM -0400
>
> You received this digest because you're subscribed to updates for this
> group. You can change your settings on the group membership page
> .
> email to cmunell+u...@googlegroups.com.
>
> You received this message because you are subscribed to the Google Groups "NELL: Never-Ending Language Learner" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to cmunell+u...@googlegroups.com.
> For more options, visit https://groups.google.com/d/optout.
>