[lojban] Re: "la" rule

10 views
Skip to first unread message

Robin Lee Powell

unread,
Nov 13, 2006, 8:06:02 PM11/13/06
to lojba...@lojban.org
On Tue, Nov 14, 2006 at 06:54:46AM +0600, Yanis Batura wrote:
> On 13.11.2006, 23:12, Robin Lee Powell wrote:
>
> > On Mon, Nov 13, 2006 at 07:08:42PM +0600, Yanis Batura wrote:
> >> On 13.11.2006, 18:47, Morphem...@wmconnect.com wrote:
> >>
> >> I would recommend a simple yet easy to learn solution that
> >> won't distort the design of the language: saying {lua} for any
> >> occurrence of {la} in cmene, {luai} for {lai} and {duoi} for
> >> {doi}.
>
> > That assumes people are capable of noticing the presence of
> > those syllables, when clearly we're not.
>
> Interesting. Will you generalize to the point that humans cannot
> detect *any* given syllables in a stream of words they're just
> going to pronounce, even after a training? Do you think there is a
> technical limitation of human brains causing this?

No, and no. I suspect that it might be *possible* to train human
brains to do the right thing here, but then it's possible to train
people to scream in terror every time you say "blue", if you try
hard enough.

I simply assert that it has shown to be so incredibly hard as to be
not worth our trouble. That is, it appears to be untrainable in
practical terms.

-Robin

--
http://www.digitalkingdom.org/~rlpowell/ *** http://www.lojban.org/
Reason #237 To Learn Lojban: "Homonyms: Their Grate!"
Proud Supporter of the Singularity Institute - http://singinst.org/


To unsubscribe from this list, send mail to lojban-li...@lojban.org
with the subject unsubscribe, or go to http://www.lojban.org/lsg2/, or if
you're really stuck, send mail to secr...@lojban.org for help.

Bob LeChevalier

unread,
Nov 13, 2006, 4:32:22 PM11/13/06
to loj...@yahoogroups.com
John E Clifford wrote:
(I don't think the {lAItl} can be read as a sumti {lai tl}).

Yes it can.

la steven. lai tl. klama

could mean that Steven is going to the Tulls.

Reminding you of the olden days, my criticism of JCB's rules focused on
the mathematician "LaPlace" (this was mentioned in my review of the last
edition of Loglan 1)

lojbab

Mark E. Shoulson

unread,
Nov 14, 2006, 11:10:19 PM11/14/06
to lojba...@lojban.org
Bob LeChevalier wrote:

> John E Clifford wrote:
>
>> It seems to me that no trick so far discussed will work in practice:
>> we will not remember to
>> exempt certain syllables from names, we will forget pauses (though
>> making {la} and the like to be
>> learned as {la.} where the period is a genuine glottal stop might
>> improve things).
>
> That is the essence of the problem.
>
> Without a solution that unquestionably will work and be used in
> practice, the justification for changing the baselined status quo
> isn't there.
Well, that's the question to be asked, isn't it. We have a solution
already, which unquestionably does NOT work. Do we jettison it in favor
of one that might? We have to pick our catastrophe.
> Those who like the pause-all-the-time solution can implement the
> practice of pausing all the time to show that in fact people can and
> will learn to do so, which could at least partially negate this
> argument; that is a legal dialect. I suspect such a dialect would be
> aesthetically displeasing, but we'd be able to judge by example.
I don't deny the potential aesthetic problems, though aesthetics is in
the ear of the beholder. A lot of Lojban's sound-system is already
considered "aesthetically displeasing" by many listeners (usually
outside of the Lojban camp). But audio-visual isomorphism is a more
important consideration for Lojban (to me) than aesthetics.
>> Again, using
>> something that occurs nowhere else (the glottal stop is a
>> difficult-to-use-or-remember-or-hear
>> example) seems the safest route (and can be used, as was pointed out
>> some time ago, to deal with
>> borrowings as well).
>
> We tried the "something that occurs nowhere else" trick for fu'ivla
> compounding, using ",iy," as the hyphen. I think someone eventually
> came up with a "somewhere else" and it was aesthetically displeasing
> to people as well.
Look, we already ARE using something that occurs nowhere else in the
word: the pause! Plans to use something even more obtrusive in that
role aren't going to help.

~mark

Robin Lee Powell

unread,
Nov 13, 2006, 8:07:26 PM11/13/06
to loj...@yahoogroups.com
On Mon, Nov 13, 2006 at 08:02:28PM -0500, Robert McIvor wrote:
>
> On 13 Nov, 2006, at 19:49, Robin Lee Powell wrote:
>
> >On Mon, Nov 13, 2006 at 07:45:21PM -0500, Robert McIvor wrote:
> >>
> >>On 13 Nov, 2006, at 17:36, Mark E. Shoulson wrote:
> >>
> >>>Robert McIvor wrote:
> >>>
> >>>>
> >>>>And the rule was altered to cover the LaPlace situation. I
> >>>>believe the current solution allows unambiguous parsing of
> >>>>names (in Loglan), and has only the restriction that names
> >>>>end with a consonant and a pause.
> >>>I'm still not fully understanding how the new Loglan rule
> >>>covers all cases. What about someone named, say {parmalagan}?
> >>>Is {la parmalagan} a single person, or is it {la parma la gan},
> >>>two sumti, the first being LA on a hypothetical predicate
> >>>{parma}?
> >>
> >>la parmalagan would be la Parmalagan, because la parma would
> >>need to be followed by a pause to be la parma la gan
> >
> >Why is that, exactly? That wasn't mentioned in your full
> >description.
>
> There has to be a pause after a name, whether a predicate used as
> a name or a proper name. This signals the end of the name. The
> pause is before the second la. I presume when a predicate is
> used with la, it is intended to specify a particular instance of
> that predicate.

Aaaaah. You didn't mention that part. That makes a huge
difference.

-Robin

--
http://www.digitalkingdom.org/~rlpowell/ *** http://www.lojban.org/
Reason #237 To Learn Lojban: "Homonyms: Their Grate!"
Proud Supporter of the Singularity Institute - http://singinst.org/

Robin Lee Powell

unread,
Nov 13, 2006, 8:33:39 PM11/13/06
to lojba...@lojban.org
On Mon, Nov 13, 2006 at 08:29:32PM -0500, Bob LeChevalier wrote:
> The la rule isn't required with standard (spaces included)
> orthography, and without a speaking community. When we have a
> skilled speaking community, we can find out whether la in names in
> the spoken language causes problems for humans - after all, humans
> can parse English and other human languages successfully, and they
> aren't audiovisually isomorphic.

So it's OK to leave a rule that people will often violate because
people can deal with that sort of thing?

That's just... It's like you *want* to kill Lojban.

I want a language that will *actually* have audio-visual
isomorphism. In practice. With real people.

You, if the above is to be believed, don't actually care about that.

Bob LeChevalier

unread,
Nov 13, 2006, 4:06:07 PM11/13/06
to lojba...@lojban.org
Seth Gordon wrote:
> Bob LeChevalier wrote:
>
>>ONLY if a word is in cmene form is it illegal to have a "la". Using a
>>gismu or lujvo or a fu'ivla as a name ("la latmo" for "Latin") has no
>>restriction.
>>
>>Indeed, I would imagine that native Lojbanists would indeed use brivla
>>to name things, rather than Lojbanize foreign words. That is indeed how
>>several languages work.
>
>
> The tutorial and reference material on Lojban that I've seen don't say a
> lot about this option. E.g., _lojban. bei loi co'a cilre_ suggests that
> since "Mei Li" is Chinese for "beautiful", then someone named Mei Li
> could Lojbanize her name as "la melb." But why not "la melbi"?
>
> Perhaps the authors of all these works just assumed that they didn't
> need to spell out this aspect of the grammar because it was so obvious,
> but it wasn't obvious to me....

It wasn't considered a "beginner" feature for English language speakers
learning the language, and all the teaching materials devised so far
have been written by at-best-intermediate speakers for the benefit of
beginners.

The potential grammar of vocatives (things that can follow "doi" or
"mi'e") is non-trivial. Likewise, sumti grammar can get quite
complicated, since we designed things to be generically compatible with
every feature of every language that we could fit in, so as not to
unnecessarily metaphysically biased in the grammar.

Even CLL, which tries to explain a lot of non-beginner features of the
language was limited by the capabilities of John, myself, and a few
others to come up with examples and usages wherein these features were used.

How much of the possibilities will ever see actual usage, we can only
guess; we can't easily teach what we barely understood how to define,
and not how and when to use ourselves.

lojbab

Robert McIvor

unread,
Nov 13, 2006, 8:29:14 PM11/13/06
to loj...@yahoogroups.com

On 13 Nov, 2006, at 20:06, Robert McIvor wrote:

I believe I remember that 'coi' identifies the author of something.
In Loglan the
canonical form would correspond to Coi La Bab, where Bab is the name.
However, a sort of usage sprung up which used the equivalent of Coi Bab,
if Bab himself was the author and Coi la Bab if one were citing a
third party.
The equivalent of 'coi is not a name marker in Loglan, so the 'la'
would be
considered the marker to be stripped

>
> On 13 Nov, 2006, at 19:52, Robin Lee Powell wrote:
>
>> On Mon, Nov 13, 2006 at 04:45:24PM -0800, Robin Lee Powell wrote:
>>> On Mon, Nov 13, 2006 at 04:41:32PM -0800, Robin Lee Powell wrote:
>>>> On Mon, Nov 13, 2006 at 07:30:47PM -0500, Robert McIvor wrote:
>>>>>
>>>>> I will assume you wish to have a rule that does not require
>>>>> recognizing the presence of the 'forbidden' combinations in
>>>>> the word. To do this, one would have to have all cmene
>>>>> marked with one of the name introducers like doi or la (I
>>>>> presume 'doi' is normally used like 'hoi' in Loglan to
>>>>> precede a name used as a vocative.
>>>
>>> Yes.
>>>
>>>>> If this be true, then the parser can strip off the
>>>>> introducer, and the remainder up to the consonant and pause
>>>>> is a name (I presume that a person without a speech defect
>>>>> would not pause in the middle of a name). The LaPlace
>>>>> problem was in sequential names. For sequential names we used
>>>>> the Loglan word for hyphen 'ci', which added 'ci' to the list
>>>>> of name markers.
>>>>
>>>> You know, I think that actually works. Or, at least, I can't
>>>> think of any problems off the top of my head.
>>>>
>>>> Call the Lojban name hyphen xi'i; laSTIvn.xi'iLAItl. has no
>>>> ambiguity I can see.
>>>
>>> However, {doilaSTIvn.} is ambiguous still. It's not a
>>> particularily hard ambiguity to fix (strip *all* name markers off
>>> the front), but still. Does Loglan allow that construct?
>>
>> Replying to myself, this is a bigger problem than I thought. Is
>> {coi.lanam.} == {coi la nam} or {coi lanam}? Does Loglan have this
>> sort of construct?
>
> I am not sure what is meant by 'coi' so I cannot answer this. If
> it were our Hoi, it would be hoi lanam.


>>
>> -Robin
>>
>> --
>> http://www.digitalkingdom.org/~rlpowell/ *** http://www.lojban.org/
>> Reason #237 To Learn Lojban: "Homonyms: Their Grate!"
>> Proud Supporter of the Singularity Institute - http://singinst.org/
>>
>>

>> To unsubscribe from this list, send mail to lojban-list-
>> req...@lojban.org


>> with the subject unsubscribe, or go to http://www.lojban.org/
>> lsg2/, or if
>> you're really stuck, send mail to secr...@lojban.org for help.
>>
>
>
>

> To unsubscribe from this list, send mail to lojban-list-
> req...@lojban.org

Seth Gordon

unread,
Nov 13, 2006, 1:44:26 PM11/13/06
to lojba...@lojban.org
Bob LeChevalier wrote:
> ONLY if a word is in cmene form is it illegal to have a "la". Using a
> gismu or lujvo or a fu'ivla as a name ("la latmo" for "Latin") has no
> restriction.
>
> Indeed, I would imagine that native Lojbanists would indeed use brivla
> to name things, rather than Lojbanize foreign words. That is indeed how
> several languages work.

The tutorial and reference material on Lojban that I've seen don't say a
lot about this option. E.g., _lojban. bei loi co'a cilre_ suggests that
since "Mei Li" is Chinese for "beautiful", then someone named Mei Li
could Lojbanize her name as "la melb." But why not "la melbi"?

Perhaps the authors of all these works just assumed that they didn't
need to spell out this aspect of the grammar because it was so obvious,
but it wasn't obvious to me....

Ryan Keppel

unread,
Nov 13, 2006, 6:23:29 PM11/13/06
to loj...@yahoogroups.com
Actually, I'll need access to several places. I happen to have a bit of time on
my hands, and intend to contribute. As I've said, my contributions will of both
time and money. I am very serious about Lojban.

.io sai

Mark E. Shoulson

unread,
Nov 13, 2006, 5:39:57 PM11/13/06
to lojba...@lojban.org
Bob LeChevalier wrote:

>
> For all that Mark thinks that the problem will be lasting, it should
> be trivial for the tools - lexer/parser/glosser etc, to include such
> things as a check for "la" and related strings in what otherwise
> parses as a cmene token, and flag it with a clear error message. In
> the absence of skilled editors, the classic standard of evaluating
> Lojban text for publication has been to run it through the parser.
"You must have faith in the future, young one. Soon, everyone will
always have computers to check their grammar."

I should put "won't this be solved by better technology?" as one of the
responses on the wiki page.

~mark

Robin Lee Powell

unread,
Nov 13, 2006, 5:03:00 PM11/13/06
to lojba...@lojban.org
On Mon, Nov 13, 2006 at 12:48:05PM -0800, John E Clifford wrote:
> (I don't think the {lAItl} can be read as a sumti {lai tl}).

Of course it can.

> > > Clearly, humans sort these out, shouldn't the parser do so as
> > > well?
> >
> > ...
> >
> > Short answer: "No; machine recognition of ambiguous languages is
> > AI-hard."
>
> But here the range of ambiguity is so small as to allow a short
> dirty run through alternatives -- or so it seems intuitively.

If we just allow la willy-nilly, there is *no way* for a parser to
tell if a speaker who says {laSTIvn.LAItl} means {la stivn laitl} or
{la stivn lai tl}, except for knowing the semantics of the language.
With the current rule, only the latter is correct.

-Robin

--
http://www.digitalkingdom.org/~rlpowell/ *** http://www.lojban.org/
Reason #237 To Learn Lojban: "Homonyms: Their Grate!"
Proud Supporter of the Singularity Institute - http://singinst.org/

Mark E. Shoulson

unread,
Nov 13, 2006, 1:13:23 PM11/13/06
to lojba...@lojban.org
Robin Lee Powell wrote:

> On Mon, Nov 13, 2006 at 07:08:42PM +0600, Yanis Batura wrote:
>
>> On 13.11.2006, 18:47, Morphem...@wmconnect.com wrote:
>>
>> I would recommend a simple yet easy to learn solution that won't
>> distort the design of the language: saying {lua} for any
>> occurrence of {la} in cmene, {luai} for {lai} and {duoi} for
>> {doi}.
>>
>
> That assumes people are capable of noticing the presence of those
> syllables, when clearly we're not.
>

If we could discern the problem cmene as we used them, we wouldn't be in
this mess. We would substitute LY or something like the CLL suggests.
Any proposed solution that involves treating "la"-containing cmene
differently from others is no better than the status quo.

Arguments for keeping the status quo seem to go along the lines of "it
will be better in the future... You must have faith, young one. Future
Lojbanists will not have this difficulty." Yeah, I don't buy it. Future
Lojbanists WILL be coining cmene on the fly, in print and in speech, for
exactly the same reasons we are now. Appeals to the salvation of the
murky future also don't hold much water for me.

It would be Just Great if there were an alternative that didn't involve
such a big change to the language, or any better alternative to the Dot
Rule. I just can't find one. The status quo is broken.

> For me, the possibility of saying a consonant-starting cmene after
> > {la} without glottal stop overweighs all disadvantages of the
> > rule. The matter of taste, of course.
That is precisely why the "la" rule was instituted in the first place.
There was a desire to speak names without preceding them by a pause, so
"guard" syllables were chosen, which thus had to be excluded from the
names. A good idea... but it apparently doesn't work.

~mark

Betsemes

unread,
Nov 15, 2006, 7:17:56 AM11/15/06
to lojba...@lojban.org
zo'o le vi nu casnu cu zdile vau u'isai

Timothy Hobbs

unread,
Nov 13, 2006, 6:06:01 PM11/13/06
to lojba...@lojban.org
what is the difference, with xorexes, between la parma and lo parma? If there are non, then gismu should not follow la, so that problem would not occur.

Ryan Keppel

unread,
Nov 13, 2006, 8:30:39 PM11/13/06
to lojba...@lojban.org
.io nai ru'e .i zoi gy. You obviously don't like me. I frankly feel the same to
you. You still haven't moved the jbosnu to a mailing list. That must take all
of a few seconds to do. You have all these completely unused lists, but no list
for lojban. I'm getting a lot of English spam on the lojban lists (la la la la
la la WHATEVER). Oh, and I donated $49. I don't want to be listed anywhere. I
just want to help Lojban. .gy.
se'a cai
fe'o

Pierre Abbat

unread,
Nov 14, 2006, 12:30:25 AM11/14/06
to lojba...@lojban.org
Robin Lee Powell wrote:
> Not without error all the time, but without error more than, say,
> 50% of the time they Lojbanize names with those sounds in them.
> Again, we have no evidence that that goal is acheivable, because we
> certainly haven't acheived it so far (3 Lojbanists reviewing WIL?
> puts us at an *upper bound* of a 33% chance of getting it right).
>
> While I was typing this mail, someone Lojbanized {malak} on #lojban,
> by the way.

People have made the error the other way too. I was looking at Wikipedia
and saw an article [[vlydimir.lenin]]. If the "la", "lai", "la'i", or
"doi" is preceded by a consonant, its presence does not invalidate the
name, e.g. {blabruk}.

As to {malak}, the part after "la" is less than a syllable. Probably he
didn't realize that you have to break everywhere a vowel is followed by
a consonant, including the last one. A single consonant is a valid
cmevla, such as {r}, the "god of hesitation". That would explain
overlooking {laitl} also; {tl} has no vowel. But then neither does {clsn}.

The "la" rule is still a lot simpler than validating fu'ivla, which
requires doing slinku'i tests and checking for lujvo.

phma

Robin Lee Powell

unread,
Nov 13, 2006, 11:20:19 PM11/13/06
to lojba...@lojban.org
On Mon, Nov 13, 2006 at 05:33:39PM -0800, Robin Lee Powell wrote:
> On Mon, Nov 13, 2006 at 08:29:32PM -0500, Bob LeChevalier wrote:
> > The la rule isn't required with standard (spaces included)
> > orthography, and without a speaking community. When we have a
> > skilled speaking community, we can find out whether la in names
> > in the spoken language causes problems for humans - after all,
> > humans can parse English and other human languages successfully,
> > and they aren't audiovisually isomorphic.
>
> So it's OK to leave a rule that people will often violate because
> people can deal with that sort of thing?
>
> That's just... It's like you *want* to kill Lojban.

I'm clearly stressing over this too much. I'm bowing out now; we
can fight about this at jbonunsla. :-)

Robert McIvor

unread,
Nov 13, 2006, 8:02:28 PM11/13/06
to lojba...@lojban.org

On 13 Nov, 2006, at 19:49, Robin Lee Powell wrote:

> On Mon, Nov 13, 2006 at 07:45:21PM -0500, Robert McIvor wrote:
>>
>> On 13 Nov, 2006, at 17:36, Mark E. Shoulson wrote:
>>
>>> Robert McIvor wrote:
>>>
>>>>
>>>> And the rule was altered to cover the LaPlace situation. I
>>>> believe the current solution allows unambiguous parsing of
>>>> names (in Loglan), and has only the restriction that names end
>>>> with a consonant and a pause.
>>> I'm still not fully understanding how the new Loglan rule covers
>>> all cases. What about someone named, say {parmalagan}? Is {la
>>> parmalagan} a single person, or is it {la parma la gan}, two
>>> sumti, the first being LA on a hypothetical predicate {parma}?
>>
>> la parmalagan would be la Parmalagan, because la parma would need
>> to be followed by a pause to be la parma la gan
>
> Why is that, exactly? That wasn't mentioned in your full
> description.

There has to be a pause after a name, whether a predicate used as a
name or a proper name. This signals the end of the name. The pause
is before the second la. I presume when a predicate is used with la,
it is intended to specify a particular instance of that predicate.
>

> Is that a pause before la, or a pause after anything after la, even
> if it's not a name?


>
> -Robin
>
> --
> http://www.digitalkingdom.org/~rlpowell/ *** http://www.lojban.org/
> Reason #237 To Learn Lojban: "Homonyms: Their Grate!"
> Proud Supporter of the Singularity Institute - http://singinst.org/
>
>

> To unsubscribe from this list, send mail to lojban-list-
> req...@lojban.org

Bob LeChevalier

unread,
Nov 13, 2006, 10:43:20 AM11/13/06
to lojba...@lojban.org
Morphem...@wmconnect.com wrote:
> Here's another reason the "la" rule is bad.
> As I understand it, ordinary gismu and lujvo can be used as names after
> "la". There are 23 gismu beginning with "la-", 16 rafsi beginning with
> "la"+consonant, 5 rafsi with "la'V" and 2 rafsi "laV". This means there
> are a lot of actual and potential words which act differently from any
> other native Lojban words, namely by becoming illegal when used as a
> cmene. For this reason alone the "la" rule should be replaced.

ONLY if a word is in cmene form is it illegal to have a "la". Using a
gismu or lujvo or a fu'ivla as a name ("la latmo" for "Latin") has no
restriction.

Indeed, I would imagine that native Lojbanists would indeed use brivla
to name things, rather than Lojbanize foreign words. That is indeed how
several languages work.

My daughter's middle name was chosen to be a valid Lojban name as well
as English name: la ka trina (Katrina)


They might even decide to Lojbanize foreign names by turning them into
type III fu'ivla (and eventually type IV if much used), which have no la
restriction.

la jecr,nortkarolaina

anyone?

mi'e la jbozbabu no'u zo'o la lojbab
(or mi'e no'u la lojbab. jbozbabu)

Robin Lee Powell

unread,
Nov 13, 2006, 7:41:32 PM11/13/06
to lojba...@lojban.org
On Mon, Nov 13, 2006 at 07:30:47PM -0500, Robert McIvor wrote:
>
> I will assume you wish to have a rule that does not require
> recognizing the presence of the 'forbidden' combinations in the
> word. To do this, one would have to have all cmene marked with
> one of the name introducers like doi or la (I presume 'doi' is
> normally used like 'hoi' in Loglan to precede a name used as a
> vocative. If this be true, then the parser can strip off the

> introducer, and the remainder up to the consonant and pause is a
> name (I presume that a person without a speech defect would not
> pause in the middle of a name). The LaPlace problem was in
> sequential names. For sequential names we used the Loglan word
> for hyphen 'ci', which added 'ci' to the list of name markers.

You know, I think that actually works. Or, at least, I can't think
of any problems off the top of my head.

Call the Lojban name hyphen xi'i; laSTIvn.xi'iLAItl. has no
ambiguity I can see.

> While writing this, I realized there may be a problem in the
> generalization. The Loglan rules were intended for the situation
> where the names were recognized as having the problem. The
> stripping off of the name marker would only occur where another
> name marker was at the beginning of a name, so the remainder
> could be recognized as a name.

I didn't follow that at all; can you give an example or something?

> Whether this would work in general would depend on whether the
> grammar allowed for a predicate to follow a name in an argument
> e.g. le la Name predicate cu .... and the predicate happened to
> have a name marker as the first syllable e.g. Loglan cibra =
> bridge, I will have to consider the general case in more detail.

That example causes no problem because "bra" can't be a cmene.

-Robin

--
http://www.digitalkingdom.org/~rlpowell/ *** http://www.lojban.org/
Reason #237 To Learn Lojban: "Homonyms: Their Grate!"
Proud Supporter of the Singularity Institute - http://singinst.org/

Bob LeChevalier

unread,
Nov 13, 2006, 8:57:17 PM11/13/06
to lojba...@lojban.org
Robert McIvor wrote:
> On 13 Nov, 2006, at 19:49, Robin Lee Powell wrote:
>
>> On Mon, Nov 13, 2006 at 07:45:21PM -0500, Robert McIvor wrote:
>>
>>>
>>> On 13 Nov, 2006, at 17:36, Mark E. Shoulson wrote:
>>>
>>>> Robert McIvor wrote:
>>>>
>>>>>
>>>>> And the rule was altered to cover the LaPlace situation. I
>>>>> believe the current solution allows unambiguous parsing of
>>>>> names (in Loglan), and has only the restriction that names end
>>>>> with a consonant and a pause.
>>>>
>>>> I'm still not fully understanding how the new Loglan rule covers
>>>> all cases. What about someone named, say {parmalagan}? Is {la
>>>> parmalagan} a single person, or is it {la parma la gan}, two
>>>> sumti, the first being LA on a hypothetical predicate {parma}?
>>>
>>>
>>> la parmalagan would be la Parmalagan, because la parma would need
>>> to be followed by a pause to be la parma la gan
>>
>>
>> Why is that, exactly? That wasn't mentioned in your full
>> description.
>
>
> There has to be a pause after a name, whether a predicate used as a
> name or a proper name. This signals the end of the name. The pause is
> before the second la. I presume when a predicate is used with la, it
> is intended to specify a particular instance of that predicate.

I believe that this is an example of what JCB called a "lexemic pause",
where the pause has grammatical import as well as lexical import (in
this case telling you that la parma la gan is two names. But of course
"parma" would be detected by a lexer as a brivla and the pause is thus
not part of "parma" but a separate lexeme. In Lojban names, the pause
at the end IS part of the cmene.

This then causes a different sort of ambiguity, between a lexemic pause,
and a hesitation or breath-pause that is otherwise legal between words.

lojbab

Chris Capel

unread,
Nov 16, 2006, 8:25:44 AM11/16/06
to lojba...@lojban.org
On 11/15/06, And Rosta <and....@gmail.com> wrote:
> Chris Capel, On 15/11/2006 03:55:
> > I have a bit of difficulty pronouncing a glottal stop in the middle of
> > a speech stream, and getting it to sound different from without, or
> > from being apparent enough. For instance, the two
> >
> > la.clsn
> > laclsn
> >
> > sound pretty much the same the way I say them.
>
> In my English English accent, _lash_ is [laS] and _latch_ is [la?S], so one can at least say that the contrast is not exotic. In Lojban, where /./ appears in difficult clusters you could insert the buffer vowel (I use [Y]), so e.g. /la.clsn/ could be [la?YSlsn]. That's what I tend to do with the /./ at the end of cmevla, e.g. {la .alis. cu} [la?alis?YSu].

I wonder whether your pronunciation is very different from mine. The
way I say "latch", I would fear using it as a pronunciation method in
Lojban because of the possible confusion with the "tc" cluster, and
ending up with something sound like "la tclsn" instead of "la .clsn".
The difference between a glottal stop and a consonant that contains a
complete airway restriction isn't apparent to me. For instance,
besides the final "n", what would the difference be between "la tcidu"
and "la .cidun"? Maybe a recording would make the difference plain
enough, though.

Chris Capel
--
"What is it like to be a bat? What is it like to bat a bee? What is it
like to be a bee being batted? What is it like to be a batted bee?"
-- The Mind's I (Hofstadter, Dennet)

Robin Lee Powell

unread,
Nov 13, 2006, 8:01:01 PM11/13/06
to loj...@yahoogroups.com
On Mon, Nov 13, 2006 at 07:55:52PM -0500, Robert McIvor wrote:
>
> On 13 Nov, 2006, at 19:45, Robin Lee Powell wrote:

>
> >On Mon, Nov 13, 2006 at 04:41:32PM -0800, Robin Lee Powell wrote:
> >>On Mon, Nov 13, 2006 at 07:30:47PM -0500, Robert McIvor wrote:
> >>>
> >>>I will assume you wish to have a rule that does not require
> >>>recognizing the presence of the 'forbidden' combinations in the
> >>>word. To do this, one would have to have all cmene marked
> >>>with one of the name introducers like doi or la (I presume
> >>>'doi' is normally used like 'hoi' in Loglan to precede a name
> >>>used as a vocative.
> >
> >Yes.

> >
> >>>If this be true, then the parser can strip off the introducer,
> >>>and the remainder up to the consonant and pause is a name (I
> >>>presume that a person without a speech defect would not pause
> >>>in the middle of a name). The LaPlace problem was in
> >>>sequential names. For sequential names we used the Loglan word
> >>>for hyphen 'ci', which added 'ci' to the list of name markers.
> >>
> >>You know, I think that actually works. Or, at least, I can't
> >>think of any problems off the top of my head.
> >>
> >>Call the Lojban name hyphen xi'i; laSTIvn.xi'iLAItl. has no
> >>ambiguity I can see.
> >
> >However, {doilaSTIvn.} is ambiguous still. It's not a
> >particularily hard ambiguity to fix (strip *all* name markers off
> >the front), but still. Does Loglan allow that construct?
>
> No. In Loglan Hoi would be used only before a bare name (which
> could be a predicate or series of predicates e.g O Wild West
> Wind. What is the purpose of doi before an argument?

I guess that that feature is to save people having to remember to
drop the {la} when greeting someone. I don't actually know.

Robert McIvor

unread,
Nov 13, 2006, 8:06:15 PM11/13/06
to lojba...@lojban.org

On 13 Nov, 2006, at 19:52, Robin Lee Powell wrote:

> On Mon, Nov 13, 2006 at 04:45:24PM -0800, Robin Lee Powell wrote:
>> On Mon, Nov 13, 2006 at 04:41:32PM -0800, Robin Lee Powell wrote:

>>> On Mon, Nov 13, 2006 at 07:30:47PM -0500, Robert McIvor wrote:
>>>>
>>>> I will assume you wish to have a rule that does not require
>>>> recognizing the presence of the 'forbidden' combinations in
>>>> the word. To do this, one would have to have all cmene
>>>> marked with one of the name introducers like doi or la (I
>>>> presume 'doi' is normally used like 'hoi' in Loglan to
>>>> precede a name used as a vocative.
>>

>> Yes.


>>
>>>> If this be true, then the parser can strip off the
>>>> introducer, and the remainder up to the consonant and pause
>>>> is a name (I presume that a person without a speech defect
>>>> would not pause in the middle of a name). The LaPlace
>>>> problem was in sequential names. For sequential names we used
>>>> the Loglan word for hyphen 'ci', which added 'ci' to the list
>>>> of name markers.
>>>
>>> You know, I think that actually works. Or, at least, I can't
>>> think of any problems off the top of my head.
>>>
>>> Call the Lojban name hyphen xi'i; laSTIvn.xi'iLAItl. has no
>>> ambiguity I can see.
>>

>> However, {doilaSTIvn.} is ambiguous still. It's not a
>> particularily hard ambiguity to fix (strip *all* name markers off
>> the front), but still. Does Loglan allow that construct?
>

> Replying to myself, this is a bigger problem than I thought. Is
> {coi.lanam.} == {coi la nam} or {coi lanam}? Does Loglan have this
> sort of construct?

I am not sure what is meant by 'coi' so I cannot answer this. If it
were our Hoi, it would be hoi lanam.
>

> -Robin
>
> --
> http://www.digitalkingdom.org/~rlpowell/ *** http://www.lojban.org/
> Reason #237 To Learn Lojban: "Homonyms: Their Grate!"
> Proud Supporter of the Singularity Institute - http://singinst.org/
>
>

> To unsubscribe from this list, send mail to lojban-list-
> req...@lojban.org

Bob LeChevalier

unread,
Nov 13, 2006, 9:26:59 PM11/13/06
to lojba...@lojban.org
Mark E. Shoulson wrote:
>> The la rule isn't required with standard (spaces included)
>> orthography, and without a speaking community. When we have a skilled
>> speaking community, we can find out whether la in names in the spoken
>> language causes problems for humans - after all, humans can parse
>> English and other human languages successfully, and they aren't
>> audiovisually isomorphic.
>
> Then why are we bothering in the first place? I thought there was a
> reason we wanted audiovisual isomorphism for Lojban.

Because it was a fundamental design requirement for Loglan. Such
fundamentals were simply not open for questioning when I redesigned the
language.

And later because audio-visual isomorphism was supposed to make the
language easier for machine speech recognition. But not because it was
a necessary principle for a human-understandable language, because
obviously it isn't.

> If that isn't a
> requirement, then we can drop a lot of other confusing stuff from the
> grammar too.
>
> Or just stick with English. People can parse that too.

Correct.

> It's like I said, pick your disaster. Is audio-visual isomorphism an
> important part of Lojban? Important enough to keep around? Is losing
> it worth keeping pauseless LA? Apparently it is, to you. If that's the
> general feeling, fine. But do recognize what it is you're doing.

What we are all doing is trying to guess what sorts of errors people
will learn *not* to make more easily, when in fact human listeners don't
typically notice the errors when speakers make them. This is true of
mandatory pauses as well as "la". We can't quantify which is the more
common error, because the mandatory pause doesn't affect the written
language, and fluent Lojban speakers/listeners who are capable of
catching pause errors while trying to understand the content probably
don't yet exist. But the pause solution would affect ALL Lojban
sentences with names in them, whereas a "la" or "doi" error only affects
certain names.

Meanwhile, it is a change to the baselined language, and I am not
especially interested in considering changes to the baseline except as
needed to get the byfy work done.

A change here would seem not to invalidate any prior Lojban - it might
change how people lojbanize their names, but it doesn't make any
particular text wrong. It thus is something that can be changed by
Lojban speakers in the post-baseline world with no impact. No decision
is needed now. And at that point, I might not be so strongly opposed to
a change.

lojbab

Yanis Batura

unread,
Nov 13, 2006, 7:54:46 PM11/13/06
to Robin Lee Powell
On 13.11.2006, 23:12, Robin Lee Powell wrote:

> On Mon, Nov 13, 2006 at 07:08:42PM +0600, Yanis Batura wrote:
>> On 13.11.2006, 18:47, Morphem...@wmconnect.com wrote:
>>
>> I would recommend a simple yet easy to learn solution that won't
>> distort the design of the language: saying {lua} for any
>> occurrence of {la} in cmene, {luai} for {lai} and {duoi} for
>> {doi}.

> That assumes people are capable of noticing the presence of those
> syllables, when clearly we're not.

Interesting. Will you generalize to the point that humans cannot


detect *any* given syllables in a stream of words they're just going
to pronounce, even after a training? Do you think there is a technical
limitation of human brains causing this?

Strange... As far as I know and judge from myself and others, human
brains can do much more difficult tasks in the realtime.

As an example, let's consider a game where players must in turns
say sentences, replacing all occurences of sound [i] with
sound, say, [u]. He who misses less [i]s, wins.

Do you think people won't get learned to play this game? Even if it
was for money?

mi'e .ianis

Timothy Hobbs

unread,
Nov 13, 2006, 3:32:04 PM11/13/06
to lojba...@lojban.org
Just a thought, but would it be possible to have mandatory terminators at the end of cmene? This terminator would be aloud at any point in the cmene, the disallowed sequence would then be much longer, it would be the terminator plus la, lo, le, ba, cu... any of the sumti or selbri starting sounds.
just a thought.

Bob LeChevalier

unread,
Nov 13, 2006, 8:29:32 PM11/13/06
to lojba...@lojban.org
Mark E. Shoulson wrote:
> Bob LeChevalier wrote:
>
>>
>> For all that Mark thinks that the problem will be lasting, it should
>> be trivial for the tools - lexer/parser/glosser etc, to include such
>> things as a check for "la" and related strings in what otherwise
>> parses as a cmene token, and flag it with a clear error message. In
>> the absence of skilled editors, the classic standard of evaluating
>> Lojban text for publication has been to run it through the parser.
>
> "You must have faith in the future, young one. Soon, everyone will
> always have computers to check their grammar."

No. Many will have computers to help them LEARN their grammar, for
those features of grammar which are necessary for computer
understanding, but which don't seem to be a problem for humans. And
this one is a very trivial check for a computer, if not for humans (who
aren't so hot even at proofreading for spelling errors in their native
language, or we wouldn't have spell-checkers).

> I should put "won't this be solved by better technology?" as one of the
> responses on the wiki page.

The la rule isn't required with standard (spaces included) orthography,

and without a speaking community. When we have a skilled speaking
community, we can find out whether la in names in the spoken language
causes problems for humans - after all, humans can parse English and
other human languages successfully, and they aren't audiovisually
isomorphic.

We know that la in names would cause problems for a computer with no AI
capability at all. So it may take a computer to be stupid enough while
rule-bound to recognize that it is misunderstanding something that
humans have no trouble understanding.

lojbab

Yanis Batura

unread,
Nov 13, 2006, 8:08:42 AM11/13/06
to lojba...@lojban.org
On 13.11.2006, 18:47, Morphem...@wmconnect.com wrote:

> Here's another reason the "la" rule is bad.
> As I understand it, ordinary gismu and lujvo can be used as names
> after "la". There are 23 gismu beginning with "la-", 16 rafsi
> beginning with "la"+consonant, 5 rafsi with "la'V" and 2 rafsi
> "laV". This means there are a lot of actual and potential words
> which act differently from any other native Lojban words, namely by
> becoming illegal when used as a cmene. For this reason alone the "la" rule should be replaced.

I would recommend a simple yet easy to learn solution that won't distort the design of the language: saying {lua} for any occurrence of {la} in cmene, {luai} for {lai} and {duoi} for {doi}.

Examples: {la luas. vegas}, {la nelson. mandeluas}.

If you want to use a gismu or lujvo in {la} construction, then add an {s}:

{larcu} -> {la luarcus}
{glare ladru} -> {la gluareluadrus}
etc.

For me, the possibility of saying a consonant-starting cmene after {la} without glottal stop overweighs all disadvantages of the rule. The matter of taste, of course.

mi'e .ianis

Bob LeChevalier

unread,
Nov 14, 2006, 4:31:32 PM11/14/06
to lojba...@lojban.org
John E Clifford wrote:
> --- Bob LeChevalier <loj...@lojban.org> wrote:
>
>>John E Clifford wrote:
>>
>>>I thought {lai} could take stress only if explicitly marked by {ba'e}. That there would also
>>
>>be
>>
>>>stress on {tl} I pass over as obvious.
>>
>>ba'e is totally independent of any sort of vocal stress.
>>
>>Why would there be stress on "tl"? It is a monosyllable. Lojban's
>>default is penultimate stress, with no stress defined for monosyllabic
>>words.
>
> Because it is a name and thus receives sentential stress and probably inherent one as well, rule
> or no rule.

I don't know what you mean by those forms of stress. Lojban has ONLY
penultimate syllable stress. No other form of stress has any semantic
or grammatical or pragmatic meaning BY DEFINITION (which is why we have
ba'e for the other kind of stress that I knew needed representation),
and no other form of stress is allowed which conflicts with the one
required form.

There is no reason in Lojban to stress a monosyllable name, or to not
stress a monosyllable name.

> It seems to me that no trick so far discussed will work in practice: we will not remember to
> exempt certain syllables from names, we will forget pauses (though making {la} and the like to be
> learned as {la.} where the period is a genuine glottal stop might improve things).

That is the essence of the problem.

Without a solution that unquestionably will work and be used in
practice, the justification for changing the baselined status quo isn't
there.

Those who like the pause-all-the-time solution can implement the
practice of pausing all the time to show that in fact people can and
will learn to do so, which could at least partially negate this
argument; that is a legal dialect. I suspect such a dialect would be
aesthetically displeasing, but we'd be able to judge by example.

> Again, using
> something that occurs nowhere else (the glottal stop is a difficult-to-use-or-remember-or-hear
> example) seems the safest route (and can be used, as was pointed out some time ago, to deal with
> borrowings as well).

We tried the "something that occurs nowhere else" trick for fu'ivla
compounding, using ",iy," as the hyphen. I think someone eventually
came up with a "somewhere else" and it was aesthetically displeasing to
people as well.

The only surefire way to do that is to mark names with a cmavo that uses
a phoneme found nowhere else in the language, including the letteral
word for the phoneme.

I wouldn't even consider trying to sell that to the community, since I
wouldn't support it myself under any conditions I can imagine.

lojbab

And Rosta

unread,
Nov 13, 2006, 10:13:44 PM11/13/06
to lojba...@lojban.org
I'm a bit mystified as to why this is provoking such debate and such arcane workarounds. Is it not the case that Mark years ago proposed changing the phonemic composition of {la, lai, doi} to {la., lai., doi.}? That change is simple and linguistically unproblematic, so the BPFK has a simple choice between introducing it or sticking with the current prescription.

--And.

Bob LeChevalier

unread,
Nov 13, 2006, 4:24:49 PM11/13/06
to lojba...@lojban.org
Timothy Hobbs wrote:
> Just a thought, but would it be possible to have mandatory
> terminators at the end of cmene? This terminator would be aloud at
> any point in the cmene, the disallowed sequence would then be much
> longer, it would be the terminator plus la, lo, le, ba, cu... any of
> the sumti or selbri starting sounds. just a thought.

The language already has that. People who are doing nonce names from
other languages probably should be doing la'o quotes, which like zoi
have a unique terminator before and after.

Learning to Lojbanize your own name is one way to learn the rules and
sound system of the language, and it is something many beginners like to
do. But there are no editors or checkers that are vetting what
self-teachers do.

For all that Mark thinks that the problem will be lasting, it should be
trivial for the tools - lexer/parser/glosser etc, to include such things
as a check for "la" and related strings in what otherwise parses as a
cmene token, and flag it with a clear error message. In the absence of
skilled editors, the classic standard of evaluating Lojban text for
publication has been to run it through the parser.

But the official parser stopped being worked on (and indeed CLL itself
wasn't run through the parser before publication, because John hadn't
kept the parser up while working on the book), and the numerous
unofficial parser projects have been reinventing the wheel in a variety
of different ways, to different ends, and haven't (to my knowledge) been
incorporating advances in the diagnostics.

But if people were regularly running their text through checkers that
caught such errors, I indeed think that people would stop making the
errors. Human beings don't parse like machines, and thus accept things
that a machine might not find legal. That just means that we need
machines to help teach us to follow some of the rules.

lojbab

John E Clifford

unread,
Nov 13, 2006, 9:53:24 PM11/13/06
to lojba...@lojban.org
Audio=visual isomorphism, isn't the problem here, but the word-isolating algorithm{
{lastIvn.lAItl}is a written, visual form, corresponding exactly (let is suppose speech competence)
the spoken (audio) utterance. The question now is, how to divide this into words and to classify
them correctly. That is also supposed to be a part of Lojban and apparently fails without either
the {la} etc prohibition or the preceding pause. I note in passing that obligatory pauses are as
likely to be skipped as {la} is to be included in names. That is, neither of these solutions
looks very practical.

--- Robin Lee Powell <rlpo...@digitalkingdom.org> wrote:

> On Mon, Nov 13, 2006 at 08:29:32PM -0500, Bob LeChevalier wrote:

> > The la rule isn't required with standard (spaces included)
> > orthography, and without a speaking community. When we have a
> > skilled speaking community, we can find out whether la in names in
> > the spoken language causes problems for humans - after all, humans
> > can parse English and other human languages successfully, and they
> > aren't audiovisually isomorphic.
>

> So it's OK to leave a rule that people will often violate because
> people can deal with that sort of thing?
>
> That's just... It's like you *want* to kill Lojban.
>

> I want a language that will *actually* have audio-visual
> isomorphism. In practice. With real people.
>
> You, if the above is to be believed, don't actually care about that.


>
> -Robin
>
> --
> http://www.digitalkingdom.org/~rlpowell/ *** http://www.lojban.org/
> Reason #237 To Learn Lojban: "Homonyms: Their Grate!"
> Proud Supporter of the Singularity Institute - http://singinst.org/
>
>

Bob LeChevalier

unread,
Nov 13, 2006, 8:11:49 PM11/13/06
to lojba...@lojban.org
Robin Lee Powell wrote:

> On Mon, Nov 13, 2006 at 04:24:49PM -0500, Bob LeChevalier wrote:
>
>>But if people were regularly running their text through checkers
>>that caught such errors, I indeed think that people would stop
>>making the errors. Human beings don't parse like machines, and
>>thus accept things that a machine might not find legal. That just
>>means that we need machines to help teach us to follow some of the
>>rules.
>
> 1. The official parser *does* accept {la.stivn.laitl.} as la
> followed by two cmene. So if you're right that all we need is
> computer training, we've already been having it for some time now,
> and it agrees with the proposed rule and not current rule.

That's what I said - the official parser had minimal sophistication for
lexing.

> 2. When you've written such a tool, let me know. Right now, you
> have no evidence that this is a human learnable rule; you're just
> having a pipe dream.

Of course it is a human-learnable rule. You underestimate human
capabilities.

The issue seems to be whether it is a rule that humans will learn to
follow *without error*, and at this point I would argue that we aren't
sure that humans can learn when elided cu and elided terminators are
valid without error. Humans don't detect those errors too easily
either, and they become important in the longer and more complex
sentences that a "native" Lojban speaker will likely use.

Adam D. Lopresto

unread,
Nov 14, 2006, 11:27:36 AM11/14/06
to lojba...@lojban.org
On Mon, 13 Nov 2006, Robin Lee Powell wrote:

> On Mon, Nov 13, 2006 at 07:55:52PM -0500, Robert McIvor wrote:
> >
> > No. In Loglan Hoi would be used only before a bare name (which
> > could be a predicate or series of predicates e.g O Wild West
> > Wind. What is the purpose of doi before an argument?
>
> I guess that that feature is to save people having to remember to
> drop the {la} when greeting someone. I don't actually know.

Well, the other purpose is that it allows any sumti (argument, whatever it is
in loglan-speak) to apply to the vocative. So we can do things like {doi la
.djan. joi la .bab.} to address several people, and use different gadri
(articles), distinguishing between {doi le finpe} (addressing the fish), {doi
lo finpe} (addressing some fish) and {doi la finpe} (addressing someone named
Fish). It's my understanding that in lojban, {doi finpe} actually acts as
shorthand for {doi le finpe}, the descriptive, and if you want to use the
predicate as a name, you need the {la}.
--
Adam Lopresto
http://cec.wustl.edu/~adam/

Press any key to continue or any other key to quit.

Robin Lee Powell

unread,
Nov 13, 2006, 3:17:35 PM11/13/06
to loj...@yahoogroups.com
On Mon, Nov 13, 2006 at 10:33:04AM -0800, John E Clifford wrote:
> Will someone please remind me what horrible things happen if we
> simply drop the restriction on cmene and make no other changes
> (that is, live as we do now -- with no reported problems). We
> might take a vocative as a sumti or, less likely, a sumti as a
> vocative.

We lose audio-visual isomorphism; la .stivn.laitl. would sound like
"la stivn lai tl".

> Clearly, humans sort these out, shouldn't the parser do
> so as well?

...

Short answer: "No; machine recognition of ambiguous languages is
AI-hard."

-Robin

--
http://www.digitalkingdom.org/~rlpowell/ *** http://www.lojban.org/
Reason #237 To Learn Lojban: "Homonyms: Their Grate!"
Proud Supporter of the Singularity Institute - http://singinst.org/

Robin Lee Powell

unread,
Nov 13, 2006, 5:23:48 PM11/13/06
to lojba...@lojban.org
On Mon, Nov 13, 2006 at 05:14:19PM -0500, Robert McIvor wrote:
>
> And the rule was altered to cover the LaPlace situation. I
> believe the current solution allows unambiguous parsing of names
> (in Loglan), and has only the restriction that names end with a
> consonant and a pause.

The rule you described didn't appear to actually solve the problem,
but then I couldn't follow it well. Can you describe it in detail for
someone (me) who doesn't know any Loglan at all?

Yanis Batura

unread,
Nov 13, 2006, 9:05:38 PM11/13/06
to lojba...@lojban.org
-----Original Message-----
From: Robin Lee Powell <rlpo...@digitalkingdom.org>
To: lojba...@lojban.org
Date: Mon, 13 Nov 2006 17:33:39 -0800
Subject: [lojban] Re: "la" rule

> So it's OK to leave a rule that people will often violate because
> people can deal with that sort of thing?
>
> That's just... It's like you *want* to kill Lojban.
>
> I want a language that will *actually* have audio-visual
> isomorphism. In practice. With real people.

I will suggest designing a special set of excersizes for remembering {la} rule for the ICSL. What I fear is that everyone except me on the ICSL list will be on the dot size. :) If it so, then... I don't know what to do, because designing excersizes using the dot rule (i.e. with {la} in cmene) will openly contradict the official grammar.

mi'e .ianis

John E Clifford

unread,
Nov 13, 2006, 3:48:05 PM11/13/06
to lojba...@lojban.org

--- Robin Lee Powell <rlpo...@digitalkingdom.org> wrote:

> On Mon, Nov 13, 2006 at 10:33:04AM -0800, John E Clifford wrote:
> > Will someone please remind me what horrible things happen if we
> > simply drop the restriction on cmene and make no other changes
> > (that is, live as we do now -- with no reported problems). We
> > might take a vocative as a sumti or, less likely, a sumti as a
> > vocative.
>
> We lose audio-visual isomorphism; la .stivn.laitl. would sound like
> "la stivn lai tl".

I don't see this one: lastIvn.lAItl. is letter perfect for the pronunciation. The problem seems,
at worst, to be that of lexing the stream: {lastIvn lAItl} may be a vocative or calling to
Lasteven or it may be a sumti (as apparently intended) (Of course, I wonder whether we remember
that obligatory pause in the middle (we don't at least as often as we do, I'd say on the basis of
limited observation). (I don't think the {lAItl} can be read as a sumti {lai tl}).



> > Clearly, humans sort these out, shouldn't the parser do
> > so as well?
>
> ...
>
> Short answer: "No; machine recognition of ambiguous languages is
> AI-hard."

But here the range of ambiguity is so small as to allow a short dirty run through alternatives --
or so it seems intuitively.

And Rosta

unread,
Nov 15, 2006, 6:41:04 AM11/15/06
to lojba...@lojban.org
Chris Capel, On 15/11/2006 03:55:
> On 11/14/06, And Rosta <and....@gmail.com> wrote:
>> The solution is
>> rather to replace the words {la}, {lai} and {doi} by {la.}, {lai.} and
>> {doi.}, i.e. [la?], [lai?], [doi?].
>
> Your "realized as a" symbol, which I assume is some IPA, shows up as a
> "?" for me. Is it a glottal stop?

Yes. Ascii IPA, not a mangled character encoding.

> I have a bit of difficulty pronouncing a glottal stop in the middle of
> a speech stream, and getting it to sound different from without, or
> from being apparent enough. For instance, the two
>
> la.clsn
> laclsn
>
> sound pretty much the same the way I say them.

In my English English accent, _lash_ is [laS] and _latch_ is [la?S], so one can at least say that the contrast is not exotic. In Lojban, where /./ appears in difficult clusters you could insert the buffer vowel (I use [Y]), so e.g. /la.clsn/ could be [la?YSlsn]. That's what I tend to do with the /./ at the end of cmevla, e.g. {la .alis. cu} [la?alis?YSu].

--And.

Mark E. Shoulson

unread,
Nov 13, 2006, 8:45:43 PM11/13/06
to lojba...@lojban.org
Bob LeChevalier wrote:

>
> The la rule isn't required with standard (spaces included)
> orthography, and without a speaking community. When we have a skilled
> speaking community, we can find out whether la in names in the spoken
> language causes problems for humans - after all, humans can parse
> English and other human languages successfully, and they aren't
> audiovisually isomorphic.

Then why are we bothering in the first place? I thought there was a

reason we wanted audiovisual isomorphism for Lojban. If that isn't a

requirement, then we can drop a lot of other confusing stuff from the
grammar too.

Or just stick with English. People can parse that too.

It's like I said, pick your disaster. Is audio-visual isomorphism an

important part of Lojban? Important enough to keep around? Is losing
it worth keeping pauseless LA? Apparently it is, to you. If that's the
general feeling, fine. But do recognize what it is you're doing.

~mark

Jorge Llambías

unread,
Nov 14, 2006, 7:20:15 AM11/14/06
to loj...@yahoogroups.com
On 11/13/06, Chris Capel <pdf...@gmail.com> wrote:
> On 11/13/06, Jorge Llamb�as <jjlla...@gmail.com> wrote:
>
> > .i stuzi la'o zoi <http://tech.groups.yahoo.com/group/jbosnu/> zoi
>
> na tolvajni toltugni djica i (but) xu djisku lu se stuzi li'u

drani

> i ta'i ma cusku zoi gy. but gy. ta'i lo ka simlu di'u

zo ku'i

mu'o mi'e xorxes

Yanis Batura

unread,
Nov 14, 2006, 12:45:26 AM11/14/06
to lojba...@lojban.org
-----Original Message-----
From: Pierre Abbat <ph...@phma.optus.nu>
To: lojba...@lojban.org
Date: Tue, 14 Nov 2006 00:30:25 -0500
Subject: [lojban] Re: "la" rule

> People have made the error the other way too. I was looking at Wikipedia
> and saw an article [[vlydimir.lenin]]. If the "la", "lai", "la'i", or
> "doi" is preceded by a consonant, its presence does not invalidate the
> name, e.g. {blabruk}.

> As to {malak}, the part after "la" is less than a syllable. Probably he
> didn't realize that you have to break everywhere a vowel is followed by
> a consonant, including the last one. A single consonant is a valid
> cmevla, such as {r}, the "god of hesitation". That would explain
> overlooking {laitl} also; {tl} has no vowel. But then neither does {clsn}.

> The "la" rule is still a lot simpler than validating fu'ivla, which
> requires doing slinku'i tests and checking for lujvo.

I strongly doubt that you will coin and verify new stage 4 fu'ivla in realtime speech. The situation is different with the {la} rule when untrained people frequently let slip those nasty {la}s to cmene.

mi'e .ianis

Robert McIvor

unread,
Nov 13, 2006, 7:45:21 PM11/13/06
to lojba...@lojban.org

On 13 Nov, 2006, at 17:36, Mark E. Shoulson wrote:

> Robert McIvor wrote:
>
>>
>> And the rule was altered to cover the LaPlace situation. I believe
>> the current solution allows unambiguous parsing of names (in
>> Loglan), and has only the restriction that names end with a
>> consonant and a pause.

> I'm still not fully understanding how the new Loglan rule covers
> all cases. What about someone named, say {parmalagan}? Is {la
> parmalagan} a single person, or is it {la parma la gan}, two sumti,
> the first being LA on a hypothetical predicate {parma}?

la parmalagan would be la Parmalagan, because la parma would need to
be followed by a pause to be la parma la gan
>

> ~mark
>
>
>
> To unsubscribe from this list, send mail to lojban-list-
> req...@lojban.org

Robert McIvor

unread,
Nov 13, 2006, 7:30:47 PM11/13/06
to lojba...@lojban.org

I will assume you wish to have a rule that does not require
recognizing the presence of the 'forbidden' combinations in the word.
To do this, one would have to have all cmene marked with one of the
name introducers like doi or la (I presume 'doi' is normally used
like 'hoi' in Loglan to precede a name used as a vocative. If this
be true, then the parser can strip off the introducer, and the
remainder up to the consonant and pause is a name (I presume that a
person without a speech defect would not pause in the middle of a
name). The LaPlace problem was in sequential names. For sequential
names we used the Loglan word for hyphen 'ci', which added 'ci' to
the list of name markers,

While writing this, I realized there may be a problem in the
generalization. The Loglan rules were intended for the situation
where the names were recognized as having the problem. The stripping
off of the name marker would only occur where another name marker was
at the beginning of a name, so the remainder could be recognized as a
name. Whether this would work in general would depend on whether the
grammar allowed for a predicate to follow a name in an argument e.g.
le la Name predicate cu .... and the predicate happened to have a
name marker as the first syllable e.g. Loglan cibra = bridge, I will
have to consider the general case in more detail.

On 13 Nov, 2006, at 17:23, Robin Lee Powell wrote:

> On Mon, Nov 13, 2006 at 05:14:19PM -0500, Robert McIvor wrote:
>>
>> And the rule was altered to cover the LaPlace situation. I
>> believe the current solution allows unambiguous parsing of names
>> (in Loglan), and has only the restriction that names end with a
>> consonant and a pause.
>
> The rule you described didn't appear to actually solve the problem,
> but then I couldn't follow it well. Can you describe it in detail for
> someone (me) who doesn't know any Loglan at all?
>
> -Robin
>
> --
> http://www.digitalkingdom.org/~rlpowell/ *** http://www.lojban.org/
> Reason #237 To Learn Lojban: "Homonyms: Their Grate!"
> Proud Supporter of the Singularity Institute - http://singinst.org/
>
>

> To unsubscribe from this list, send mail to lojban-list-
> req...@lojban.org

Robert McIvor

unread,
Nov 13, 2006, 7:55:52 PM11/13/06
to lojba...@lojban.org

On 13 Nov, 2006, at 19:45, Robin Lee Powell wrote:

> On Mon, Nov 13, 2006 at 04:41:32PM -0800, Robin Lee Powell wrote:
>> On Mon, Nov 13, 2006 at 07:30:47PM -0500, Robert McIvor wrote:
>>>

>>> I will assume you wish to have a rule that does not require
>>> recognizing the presence of the 'forbidden' combinations in the
>>> word. To do this, one would have to have all cmene marked with
>>> one of the name introducers like doi or la (I presume 'doi' is
>>> normally used like 'hoi' in Loglan to precede a name used as a
>>> vocative.
>

> Yes.


>
>>> If this be true, then the parser can strip off the introducer,
>>> and the remainder up to the consonant and pause is a name (I
>>> presume that a person without a speech defect would not pause
>>> in the middle of a name). The LaPlace problem was in
>>> sequential names. For sequential names we used the Loglan word

>>> for hyphen 'ci', which added 'ci' to the list of name markers.


>>
>> You know, I think that actually works. Or, at least, I can't
>> think of any problems off the top of my head.
>>
>> Call the Lojban name hyphen xi'i; laSTIvn.xi'iLAItl. has no
>> ambiguity I can see.
>
> However, {doilaSTIvn.} is ambiguous still. It's not a particularily
> hard ambiguity to fix (strip *all* name markers off the front), but
> still. Does Loglan allow that construct?

No. In Loglan Hoi would be used only before a bare name (which could

be a predicate or series of predicates e.g O Wild West Wind. What is
the purpose of doi before an argument?
>

Seth Gordon

unread,
Nov 13, 2006, 4:36:41 PM11/13/06
to loj...@yahoogroups.com
Bob LeChevalier wrote:
>> The tutorial and reference material on Lojban that I've seen don't say a
>> lot about this option. E.g., _lojban. bei loi co'a cilre_ suggests that
>> since "Mei Li" is Chinese for "beautiful", then someone named Mei Li
>> could Lojbanize her name as "la melb." But why not "la melbi"?
>>
>> Perhaps the authors of all these works just assumed that they didn't
>> need to spell out this aspect of the grammar because it was so obvious,
>> but it wasn't obvious to me....
>
>
> It wasn't considered a "beginner" feature for English language speakers
> learning the language, and all the teaching materials devised so far
> have been written by at-best-intermediate speakers for the benefit of
> beginners.
>
> The potential grammar of vocatives (things that can follow "doi" or
> "mi'e") is non-trivial. Likewise, sumti grammar can get quite
> complicated, since we designed things to be generically compatible with
> every feature of every language that we could fit in, so as not to
> unnecessarily metaphysically biased in the grammar.

Ah. It seemed like a "beginner" feature to me because as far as I could
glean, one could replace "le" or "lo" with "la" in a phrase, and the
phrase would still parse--it would be like substituting "Spider-Man" for
"the arachnoid human". But I suppose once "doi" and "mi'e" are thrown
into the mix then one has to be more careful.

Bob LeChevalier

unread,
Nov 13, 2006, 9:04:57 PM11/13/06
to lojba...@lojban.org
Robert McIvor wrote:
>
> On 13 Nov, 2006, at 20:06, Robert McIvor wrote:
>
> I believe I remember that 'coi' identifies the author of something. In
> Loglan the
> canonical form would correspond to Coi La Bab, where Bab is the name.
> However, a sort of usage sprung up which used the equivalent of Coi Bab,
> if Bab himself was the author and Coi la Bab if one were citing a third
> party.
> The equivalent of 'coi is not a name marker in Loglan, so the 'la'
> would be
> considered the marker to be stripped

Lojban's vocative marker is "doi", having no semantic content other than
implicit in the grammatical marking. "coi" means "hello" and has the
same grammar as the vocative marker, but the obvious semantics and
pragmatics of greeting. Lojban has several other members of the
vocative selma'o/lexeme.

lojbab

Bob LeChevalier

unread,
Nov 13, 2006, 9:45:22 PM11/13/06
to lojba...@lojban.org
John E Clifford wrote:
> I thought {lai} could take stress only if explicitly marked by {ba'e}. That there would also be
> stress on {tl} I pass over as obvious.

ba'e is totally independent of any sort of vocal stress.

Why would there be stress on "tl"? It is a monosyllable. Lojban's
default is penultimate stress, with no stress defined for monosyllabic
words.

If the last syllable of a word is stressed and the following word is a
*brivla*, a pause is mandatory. But there is no rule for words followed
by names other than a pause except after la, lai, la'i, and doi (page 69
of CLL).

The present issue was in fact alluded to on that page in CLL. It was
noted that the situations where failure to pause before a name causes
problems for a listener are rare. It is because it rarely causes
problems for listeners, that people get away with errors incorporating "la".

Robin Lee Powell

unread,
Nov 13, 2006, 8:19:49 PM11/13/06
to loj...@yahoogroups.com
On Mon, Nov 13, 2006 at 08:11:49PM -0500, Bob LeChevalier wrote:
> Robin Lee Powell wrote:
> >2. When you've written such a tool, let me know. Right now, you
> >have no evidence that this is a human learnable rule; you're just
> >having a pipe dream.
>
> Of course it is a human-learnable rule. You underestimate human
> capabilities.
>
> The issue seems to be whether it is a rule that humans will learn
> to follow *without error*,

Not without error all the time, but without error more than, say,
50% of the time they Lojbanize names with those sounds in them.
Again, we have no evidence that that goal is acheivable, because we
certainly haven't acheived it so far (3 Lojbanists reviewing WIL?
puts us at an *upper bound* of a 33% chance of getting it right).

While I was typing this mail, someone Lojbanized {malak} on #lojban,
by the way.

The rule obnoxious, it gains us nothing whatever that *I* care about
(oh no, a glottal stop! .ii sai je'u nai), and it appears to be
very hard to follow.

Again, I don't think either of us are saying anything new at this
point.

-Robin

--
http://www.digitalkingdom.org/~rlpowell/ *** http://www.lojban.org/
Reason #237 To Learn Lojban: "Homonyms: Their Grate!"
Proud Supporter of the Singularity Institute - http://singinst.org/

Chris Capel

unread,
Nov 14, 2006, 10:55:01 PM11/14/06
to loj...@yahoogroups.com
On 11/14/06, And Rosta <and....@gmail.com> wrote:
> Bob LeChevalier, On 14/11/2006 21:31:

> > Those who like the pause-all-the-time solution can implement the
> > practice of pausing all the time to show that in fact people can and
> > will learn to do so, which could at least partially negate this
> > argument; that is a legal dialect.
>
> "Pause-all-the-time" is a very misleading description. The solution is rather to replace the words {la}, {lai} and {doi} by {la.}, {lai.} and {doi.}, i.e. [la?], [lai?], [doi?].
>
> The idea that the phoneme /./ is realized by a pause at all, let alone as its primary allophone, is lunacy in a human language. A much more sensible analysis of the situation in Lojban is that the phoneme /./ (realized as [?]) can be unrealized when at the edge of a phonological string.

Your "realized as a" symbol, which I assume is some IPA, shows up as a
"?" for me. Is it a glottal stop?

I have a bit of difficulty pronouncing a glottal stop in the middle of


a speech stream, and getting it to sound different from without, or
from being apparent enough. For instance, the two

la.clsn
laclsn

sound pretty much the same the way I say them.

Chris Capel
--
"What is it like to be a bat? What is it like to bat a bee? What is it
like to be a bee being batted? What is it like to be a batted bee?"
-- The Mind's I (Hofstadter, Dennet)

Robert McIvor

unread,
Nov 13, 2006, 5:14:19 PM11/13/06
to lojba...@lojban.org

And the rule was altered to cover the LaPlace situation. I believe
the current solution allows unambiguous parsing of names (in Loglan),
and has only the restriction that names end with a consonant and a
pause.

On 13 Nov, 2006, at 16:32, Bob LeChevalier wrote:

> John E Clifford wrote:
> (I don't think the {lAItl} can be read as a sumti {lai tl}).
>

> Yes it can.
>
> la steven. lai tl. klama
>
> could mean that Steven is going to the Tulls.
>
> Reminding you of the olden days, my criticism of JCB's rules
> focused on the mathematician "LaPlace" (this was mentioned in my
> review of the last edition of Loglan 1)
>
> lojbab
>
>
>
>
> To unsubscribe from this list, send mail to lojban-list-
> req...@lojban.org

And Rosta

unread,
Nov 14, 2006, 6:19:44 PM11/14/06
to lojba...@lojban.org
Bob LeChevalier, On 14/11/2006 21:31:
> John E Clifford wrote:
>> It seems to me that no trick so far discussed will work in practice:
>> we will not remember to
>> exempt certain syllables from names, we will forget pauses (though
>> making {la} and the like to be
>> learned as {la.} where the period is a genuine glottal stop might
>> improve things).
>
> That is the essence of the problem.
>
> Without a solution that unquestionably will work and be used in
> practice, the justification for changing the baselined status quo isn't
> there.

The Clsn solution works. It's hard to see why it wouldn't get used if it became official.



> Those who like the pause-all-the-time solution can implement the
> practice of pausing all the time to show that in fact people can and
> will learn to do so, which could at least partially negate this
> argument; that is a legal dialect.

"Pause-all-the-time" is a very misleading description. The solution is rather to replace the words {la}, {lai} and {doi} by {la.}, {lai.} and {doi.}, i.e. [la?], [lai?], [doi?].

The idea that the phoneme /./ is realized by a pause at all, let alone as its primary allophone, is lunacy in a human language. A much more sensible analysis of the situation in Lojban is that the phoneme /./ (realized as [?]) can be unrealized when at the edge of a phonological string.

> I suspect such a dialect would be

> aesthetically displeasing, but we'd be able to judge by example.

The aesthetic effect would be essentially identical to the status quo.

Fair dos -- by all means argue for the sanctity of the baseline, but base that the standard sociopolitical arguments, not on a load of specious bollocks of the sort given above.

--And.

And Rosta

unread,
Nov 18, 2006, 6:44:30 PM11/18/06
to lojba...@lojban.org
Chris Capel, On 16/11/2006 13:25:

> On 11/15/06, And Rosta <and....@gmail.com> wrote:
>> Chris Capel, On 15/11/2006 03:55:
>> > la.clsn
>> > laclsn
>> >
>> > sound pretty much the same the way I say them.
>>
>> In my English English accent, _lash_ is [laS] and _latch_ is [la?S],
>> so one can at least say that the contrast is not exotic. In Lojban,
>> where /./ appears in difficult clusters you could insert the buffer
>> vowel (I use [Y]), so e.g. /la.clsn/ could be [la?YSlsn]. That's what
>> I tend to do with the /./ at the end of cmevla, e.g. {la .alis. cu}
>> [la?alis?YSu].
>
> I wonder whether your pronunciation is very different from mine. The
> way I say "latch", I would fear using it as a pronunciation method in
> Lojban because of the possible confusion with the "tc" cluster, and
> ending up with something sound like "la tclsn" instead of "la .clsn".
> The difference between a glottal stop and a consonant that contains a
> complete airway restriction isn't apparent to me. For instance,
> besides the final "n", what would the difference be between "la tcidu"
> and "la .cidun"? Maybe a recording would make the difference plain
> enough, though.

Acoustically, the contrast between Lojban /a.ci/, /atci/, /akci/ and /apci/ is manifest in the different ways the /., t, c, p/ warp the formants of the preceding /a/. But for the likes of you and me, the acoustic contrast would be hard to perceive (when not consciously listening out for it) because in English [?] and [t] are both allophones of /t/ and are sometimes in free variation. But it's inevitable that this sort of thing happens when one uses a foreign language.

Matt Arnold

unread,
Nov 13, 2006, 4:43:10 PM11/13/06
to lojba...@lojban.org
On 11/13/06, Bob LeChevalier <loj...@lojban.org> wrote:
> the numerous
> unofficial parser projects have been reinventing the wheel in a variety
> of different ways, to different ends, and haven't (to my knowledge) been
> incorporating advances in the diagnostics.
>
> But if people were regularly running their text through checkers that
> caught such errors, I indeed think that people would stop making the
> errors. Human beings don't parse like machines, and thus accept things
> that a machine might not find legal. That just means that we need
> machines to help teach us to follow some of the rules.
>
> lojbab


In my opinion, easy-to-use computer prosthetics embedded into a
Firefox extension would be a major milestone for the language. It
would be as if we had a native speaker to teach us. Admittedly, Lojban
is not a computer language, but it has unique advantages for computer
diagnostics, and therefore computers should be our teachers. We should
be able to use ubiquitous built-in Lojban parsing tools to prompt us
every time we make a mistake in Lojban.

-Eppcott

John E Clifford

unread,
Nov 13, 2006, 9:27:15 PM11/13/06
to lojba...@lojban.org
I thought {lai} could take stress only if explicitly marked by {ba'e}. That there would also be
stress on {tl} I pass over as obvious.

--- Bob LeChevalier <loj...@lojban.org> wrote:

> John E Clifford wrote:
> (I don't think the {lAItl} can be read as a sumti {lai tl}).
>
> Yes it can.
>
> la steven. lai tl. klama
>
> could mean that Steven is going to the Tulls.
>
> Reminding you of the olden days, my criticism of JCB's rules focused on
> the mathematician "LaPlace" (this was mentioned in my review of the last
> edition of Loglan 1)
>
> lojbab
>
>
>
>

Pierre Abbat

unread,
Nov 13, 2006, 10:22:22 AM11/13/06