Received-SPF: pass (google.com: best guess record for domain of mbays@sdf.org designates 192.94.73.19 as permitted sender) client-ip=192.94.73.19;
Date: Fri, 25 Nov 2011 14:50:38 -0500
From: Martin Bays <mbays@sdf.org>
To: lojban@googlegroups.com
Subject: Re: [lojban] semantic parser - tersmu-0.1rc1
Message-ID: <20111125195038.GA29512@gonzales>
References: <20111124044118.GF6112@gonzales>
 <CAO7tK2cfhJtrvLX1S6Mga_CJU2nLyt96_4DR=ygJksH3ixaamg@mail.gmail.com>
MIME-Version: 1.0
Content-Type: multipart/signed; micalg=pgp-sha1;
	protocol="application/pgp-signature"; boundary="gKMricLos+KVdGMg"
Content-Disposition: inline
In-Reply-To: <CAO7tK2cfhJtrvLX1S6Mga_CJU2nLyt96_4DR=ygJksH3ixaamg@mail.gmail.com>
X-cunselcu'a-valsi: sfofa
User-Agent: Mutt/1.5.21 (2010-09-15)
X-Original-Sender: mbays@sdf.org
X-Original-Authentication-Results: gmr-mx.google.com; spf=pass (google.com:
 best guess record for domain of mbays@sdf.org designates 192.94.73.19 as
 permitted sender) smtp.mail=mbays@sdf.org
Reply-To: lojban@googlegroups.com
Precedence: list
Mailing-list: list lojban@googlegroups.com; contact lojban+owners@googlegroups.com
List-ID: <lojban.googlegroups.com>
X-Google-Group-Id: 1004133512417
List-Post: <http://groups.google.com/group/lojban/post?hl=en_US>, <mailto:lojban@googlegroups.com>
List-Help: <http://groups.google.com/support/?hl=en_US>, <mailto:lojban+help@googlegroups.com>
List-Archive: <http://groups.google.com/group/lojban?hl=en_US>
Sender: lojban@googlegroups.com
List-Subscribe: <http://groups.google.com/group/lojban/subscribe?hl=en_US>, <mailto:lojban+subscribe@googlegroups.com>
List-Unsubscribe: <http://groups.google.com/group/lojban/subscribe?hl=en_US>, <mailto:lojban+unsubscribe@googlegroups.com>
X-Spam-Score: -0.7 (/)
X-Spam_score: -0.7
X-Spam_score_int: -6
X-Spam_bar: /


--gKMricLos+KVdGMg
Content-Type: text/plain; charset=iso-8859-1
Content-Disposition: inline
Content-Transfer-Encoding: quoted-printable

* Thursday, 2011-11-24 at 21:18 -0300 - Jorge Llamb=EDas <jjllambias@gmail.=
com>:

> On Thu, Nov 24, 2011 at 1:41 AM, Martin Bays <mbays@sdf.org> wrote:
> >
> > Probably not bugs:
> > =A0 =A0This might at first seem wrong:
> > =A0 =A0 =A0 =A0> na ku mi noi brode cu broda
> > =A0 =A0 =A0 =A0Prop:!(brode(mi) /\ broda(mi))
>=20
> It does seem wrong to me. A noi-clause is more like a presupposition,
> not directly part of the claim.
>=20
> I'd say it's more like:
>=20
> presupp Prop: brode(mi)
> main Prop: !broda(mi)

Yes, quite possibly.

I was working under the assumption that {noi} should be part of this
basic logical fragment of lojban, meaning in particular that we don't
have to break out of existing quantifiers when parsing it.

Clearly some parts of lojban do require breaking out in this way -
question words, anaphora and probably xorlo being examples - so adding
{noi} to that list wouldn't be too shocking. All else being equal,
however, I would prefer not to.

> > =A0 =A0but consider that e.g.
> > =A0 =A0 =A0 =A0> na ku da ro broda be da ku noi brodi cu brodu
> > =A0 =A0 =A0 =A0Prop:!EX x1. FA x2:(broda(_,x1)). (brodu(x1,x2) /\ brodi=
(x2))
> > =A0 =A0is probably right.
>=20
> I don't think a noi-clause is well-defined when it is attached to
> something that doesn't have referents. You need a referent in order to
> be able to make a comment about it. It may be that noi forces
> something like:
>=20
> ?presupp Prop: FA x1. FA x2:(broda(_,x1). brodi(x2)
> main Prop: !EX x1. FA x2:(broda(_,x1)). brodu(x1,x2)
>=20
> I'm not completely sure if that's exactly what the presupposition
> should be, but I do believe the main proposition should not be
> affected if you remove any noi-clause from it.

I don't know.

Consider analogous english phrases:

Not one of the farmers owns donkeys, which he beats, with long ears.

(excuse sexist pronoun; using anything else would confuse the issue)

I think the relative phrase here is, on its own, unambiguously
incidental. But the sentence does not imply that all of the farmers beat
their donkeys. The overall effect is that the incidental phrase does act
restrictively, just as in my interpretation of the corresponding lojban:

no le cange cu ponse su'o xasli noi ri darxi zi'e poi clani se kerlo

Prop: {le} x1:(cange(_)). EQ(0) x2:(me(_,x1)). EX x3:((xasli(_) /\
    <clani(_)><kerlo>( ,_))). (ponse(x2,x3) /\ darxi(x2,x3))

jbo: le cange ku goi ko'a 0 da poi ke'a me ko'a ku'o su'o de poi ge ke'a
    xasli gi zo'e ke clani kerlo ke'e ke'a ku'o zo'u ge da ponse de gi
    da darxi de


Similarly: "Each boy carried his bag. Two of the boys carried his bag,
which contained his lunch, on his head" doesn't imply that each bag
contained its owner's lunch. Again, this agrees with:

ro lo nanla goi ny cu bevri lo ri dakli .i je re ny stedu cpana bevri lo
ri dakli noi vasru lo ri sanmi=20

Prop: {lo} x1:(nanla(_)). FA x2:(me(_,x1)). {lo} x3:((srana(x2,_) /\
    dakli(_))). EQ(2) x4:(me(_,x1)). {lo} x5:((srana(x4,_) /\
    dakli(_))). (bevri(x2,x3) /\ (<<stedu(_)><cpana>(_)><bevri>(x4,x5)
    /\ {lo} x6:((srana(x4,_) /\ sanmi(_))). vasru(x5,x6)))

jbo: lo nanla ku goi ko'a ro da poi ke'a me ko'a ku'o lo poi'i ge da
    srana ke'a gi ke'a dakli kei ku goi ko'e 2 de poi ke'a me ko'a ku'o lo
    poi'i ge de srana ke'a gi ke'a dakli kei ku goi ko'i zo'u ge da bevri
    ko'e gi ge de ke ke stedu cpana ke'e bevri ke'e ko'i gi lo poi'i ge de
    srana ke'a gi ke'a sanmi kei ku goi ko'o zo'u ko'i vasru ko'o


Then again, the english "It is impossible that I, who am great, could
fail" probably isn't falsified by my possibly not being great (although
I suspect it's ambiguous). So incidentals in english probably are
sometimes scope-jumping in the way you suggest.


> > =A0 =A0Also,
> > =A0 =A0 =A0 =A0> ro da na ku broda .i je de brode
> > =A0 =A0 =A0 =A0Prop:FA x1. !EX x2. (broda(x1) /\ brode(x2))
> > =A0 =A0is right, because
> > =A0 =A0 =A0 =A0> ro da na ku broda de .i je de brode
> > =A0 =A0 =A0 =A0Prop:FA x1. !EX x2. (broda(x1,x2) /\ brode(x2))
> > =A0 =A0has to be; c.f.
> > =A0 =A0 =A0 =A0> ro da na broda de .i je de brode
> > =A0 =A0 =A0 =A0Prop:FA x1. EX x2. (!broda(x1,x2) /\ brode(x2)) .
>=20
> This is a hairy issue. My instinct has always been that ".ije" should
> be able to connect prenexed sentences, but the official grammar does
> say otherwise.

It does. I initially implemented only 'subsentence'; getting from that
to 'text' was much more difficult than I had anticipated, largely
because of this kind of issue.

But having connected sentences share a common prenex does seem often
useful, e.g. to keep a bound variable alive, and unintuitive only if you
have the wrong intuition (which I certainly did at first!).

> BTW, how do you analyse "su'o da na broda" and "su'o da na broda gi'e na =
brode"?

Prop: EX x1. !broda(x1)
and
Prop: EX x1. (!broda(x1) /\ !brode(x1))
respectively.

This current handling does mean we get possibly unintuitive results once
tailterms are brought in:

da na broda de
Prop: EX x1. EX x2. !broda(x1,x2)
jbo: su'o da su'o de zo'u na ku da broda de

su'o da na broda gi'e na brode vau de
Prop: EX x1. EX x2. (!broda(x1,x2) /\ !brode(x1,x2))
jbo: su'o da su'o de zo'u ge na ku da broda de gi na ku da brode de

The obvious alternative would be to have {na broda} work like {broda be
na ku}, which I think corresponds to what you suggest on the BPFK
section page, giving us

da broda be na ku de
Prop: EX x1. !EX x2. broda(x1,x2)
jbo: su'o da zo'u na ku su'o de zo'u da broda de

but

su'o da broda na ku gi'e brode na ku
Prop: EX x1. !!(broda(x1) /\ brode(x1))
jbo: su'o da zo'u na ku na ku ge da broda gi da brode

and

su'o da broda be na ku gi'e brode be na ku
Prop: EX x1. !!(broda(x1) /\ brode(x1))
jbo: su'o da zo'u na ku na ku ge da broda gi da brode

and

su'o da broda be na ku gi'e brode be na ku vau de
Prop: EX x1. !EX x2. !(broda(x1,x2) /\ brode(x1,x2))
jbo: su'o da zo'u na ku su'o de zo'u na ku ge da broda de gi da brode de

Contrast all that with

su'o da broda na gi'e nai brode vau de
Prop: EX x1. EX x2. (!broda(x1,x2) /\ !brode(x1,x2))
jbo: su'o da su'o de zo'u ge na ku da broda de gi na ku da brode de


I'm not sure what the correct behaviour is here (also generally with
tailterms of connected briditails, even without the negation)


By the way: in case you or anyone else wants to play with my current
implementation without compiling it yourself, you can now do so by
sshing to:

gonzales.homelinux.org  (username tersmu; password tersmu)

I won't actually be watching you to the extent that the (traditional)
motd might suggest.

Martin

--gKMricLos+KVdGMg
Content-Type: application/pgp-signature

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2.0.17 (GNU/Linux)

iEYEARECAAYFAk7P8Y4ACgkQULC7OLX7LNYtSQCeMrRVzqnrNmSSsMyXZv5I5MHo
hdYAoMFdDrkeNx1bM2LeMBP4xEWTqWvk
=zalq
-----END PGP SIGNATURE-----

--gKMricLos+KVdGMg--