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

Re: [lojban] Semantic Representation of Lojban



I've realized that I'll need to look at sentences like "I promise X
[though my finger is crossed.]"[1] because of their appearing to say
"I promise X."  And I've gotten about that far with that particular
problem.  :-)

I'm only clear on the first three steps I'll be taking to work on
this problem, and I'm roughly planning on adding bits of logic and
leftovers until I find something that appears to contradict some
step I've already taken.  Or, more likely, I'm not nearly so lucky
as that and reach a point where I don't understand what a Lojban
statement class is formally trying to express, and need help
translating it.

I'm really happy you're interested in this stuff.  I'll be happy to
check in when I have an interesting application, a fun example, or
anything else thought provoking.  If you build any datasets I'd be
happy to hear about them as well.

Happy Hacking,

-Alan

1: http://en.wikipedia.org/wiki/Crossed_fingers, though the cultural
   referent I'm making (documented in the first paragraph) is not
   cited!

On Thu, Jan 06, 2011 at 04:49:38PM -0800, John E Clifford wrote:
> OK, cool.  My thought was to do the pred log first and then deal with what is 
> left over, yours is to sort out the pred log and (much of) the leftover first 
> and deal with a lot of that leftover.  We are both going to run into the problem 
> of overlap at some point, apparently factual sentences that are used to express 
> emotions and the opposite.  Until the, however, we can start with the clearly 
> marked bits of Lojban: attitudinals to express emotions, imperatives to express 
> requests, the fraught sentences with 'djica' and 'nitcu' to express/report 
> needs.  So far, this is vocab look up and a little parsing.  What comes next in 
> terms of making these classifications when they are embedded or implicit or 
> disguised?  And, of course, how much of this hinges on fairly complete word 
> semantics for various preds and compounds?  The first part won't take a thousand 
> hours (nor probably a hundred). the next part leads you into the whirlpool that 
> also involves the other approach, so step-wise mutual adapting, hopefully always 
> in an upward direction.
> 
> 
> 
> ----- Original Message ----
> From: .alyn.post. <alyn.post@lodockikumazvati.org>
> To: lojban@googlegroups.com
> Sent: Thu, January 6, 2011 4:35:57 PM
> Subject: Re: [lojban] Semantic Representation of Lojban
> 
> I'd like to ellaborate a little more on this, because the link I
> provided has a low information content.  I need to solve problems
> like this:
> 
> Nonviolent Communication[1] has a four-part communication model
> called OFNR[2].  In it, you state your observation, your feeling,
> your need, and (optionally) a request.
> 
> This turns out to be difficult, because we constantly mix
> observations and feelings, state judgements as if they were
> feelings, &c.  It is something that takes practice to do well.
> 
> I want to be able to classify a Lojban statement based on a
> constraint, like this:
> 
>   Does statement X consist of an observation, then a feeling, then
>   a need, and then optionally a request.
> 
> The idea being that one can type in what *seems* like an OFNR
> statement, and have the computer call you on it if it isn't.
> 
> The above is actually quite complex, compared to the first version
> of what I want to do, which is to have a conversation with ~20 valsi
> and a finite state machine with something like that many states, and
> use a simple version of the above classifier as to control state
> transition.  The classifier won't need to handle anything outside
> the scope of those ~20 valsi, save to transition to a "maybe you
> ought to write that piece you clever monkey" state.
> 
> I hope that demonstrates that I don't need a formal solution to the
> problem, but that I need something *like* a solution to the problem,
> and I hope that gives a better idea of the kind of thing I have in
> mind, as I'm working on and interested in this topic.
> 
> -Alan
> 
> 1: http://www.cnvc.org/
> 2: http://en.nvcwiki.com/index.php/Four_part_model
> 
> On Thu, Jan 06, 2011 at 03:11:07PM -0700, .alyn.post. wrote:
> > By "playing with it," I mean specifically that I have an application
> > in mind[1] and will be doing the minimum amount of work in this
> > domain to support that application, with my goal being to develop
> > and deliver the application, rather than a formal solution to this
> > problem.
> > 
> > The initial version of the application requires only a pathetically
> > bad approximation to this problem, and so I will be able to use the
> > result in well under a few thousand man-hours.
> > 
> > -Alan
> > 
> > 1: http://wiki.call-cc.org/eggref/4/kiksispehi
> > 
> > On Thu, Jan 06, 2011 at 02:01:56PM -0800, John E Clifford wrote:
> > > As a practical matter, the first (and officially easiest -- but time will 
> >tell) 
> >
> > > part would be to devise the rules for working back from Lojban surface 
> > > structures to the underlying predicate logic ones: all logical connectives 
> > > between sentences, all quantifiers and negations in appropriate prenex 
> >position 
> >
> > > (so the structure that immediate follows -- a sentence of some sort -- will 
> >be 
> >
> > > exactly the intended scope).  You might also start a bunch of meaning 
> > > postulates, that relate one concept to others (I suppose, at least initially. 
> >
> > > that the metalanguage will be English) and throw in the laws of logic just in 
> >
> > > case (but they are probably going to be needed early on anyhow, to sort out 
> > > issues in prenectification).  That ought to be worth a few thousand 
> >man-hours.
> > > 
> > > 
> > > 
> > > ----- Original Message ----
> > > From: .alyn.post. <alyn.post@lodockikumazvati.org>
> > > To: lojban@googlegroups.com
> > > Sent: Thu, January 6, 2011 2:45:34 PM
> > > Subject: [lojban] Semantic Representation of Lojban
> > > 
> > > [I've moved this to it's own thread for higher visibility of the
> > > topic.]
> > > 
> > > On Thu, Jan 06, 2011 at 12:38:23PM -0800, Robin Lee Powell wrote:
> > > > > Well, to a certain extent you're right, but if you choose the
> > > > > right kind of semantic representation, you can do things like
> > > > > proving that two different strings of Lojban have the same
> > > > > meaning. Correct me if I'm wrong, but at the moment no machine
> > > > > grammar of Lojban represents the fact that “mi viska do” is
> > > > > equivalent to “do se viska mi”.
> > > > 
> > > > Right, very true.  People have started playing with that.
> > > > 
> > > 
> > > I've started playing with it, certainly.  Enough to where I'm
> > > considering flying out to Penguicon to brainstorm and talk about
> > > it with other Lojbanists.
> > > 
> > > If others of you are working on it and are able and interested in
> > > meeting about it, will you speak up?
> > > 
> > > -Alan
> > > -- 
> > > .i ko djuno fi le do sevzi
> > > 
> > > -- 
> > > You received this message because you are subscribed to the Google Groups 
> > > "lojban" group.
> > > To post to this group, send email to lojban@googlegroups.com.
> > > To unsubscribe from this group, send email to 
> > > lojban+unsubscribe@googlegroups.com.
> > > For more options, visit this group at 
> > > http://groups.google.com/group/lojban?hl=en.
> > > 
> > > 
> > >      
> > > 
> > > -- 
> > > You received this message because you are subscribed to the Google Groups 
> >"lojban" group.
> > > To post to this group, send email to lojban@googlegroups.com.
> > > To unsubscribe from this group, send email to 
> >lojban+unsubscribe@googlegroups.com.
> > > For more options, visit this group at 
> >http://groups.google.com/group/lojban?hl=en.
> > > 
> > 
> > -- 
> > .i ko djuno fi le do sevzi
> 
> -- 
> .i ko djuno fi le do sevzi
> 
> -- 
> You received this message because you are subscribed to the Google Groups 
> "lojban" group.
> To post to this group, send email to lojban@googlegroups.com.
> To unsubscribe from this group, send email to 
> lojban+unsubscribe@googlegroups.com.
> For more options, visit this group at 
> http://groups.google.com/group/lojban?hl=en.
> 
> 
>       
> 
> -- 
> You received this message because you are subscribed to the Google Groups "lojban" group.
> To post to this group, send email to lojban@googlegroups.com.
> To unsubscribe from this group, send email to lojban+unsubscribe@googlegroups.com.
> For more options, visit this group at http://groups.google.com/group/lojban?hl=en.
> 

-- 
.i ko djuno fi le do sevzi

-- 
You received this message because you are subscribed to the Google Groups "lojban" group.
To post to this group, send email to lojban@googlegroups.com.
To unsubscribe from this group, send email to lojban+unsubscribe@googlegroups.com.
For more options, visit this group at http://groups.google.com/group/lojban?hl=en.