Homonyms in Stage 3 fu'ivla

39 views
Skip to first unread message

guskant

unread,
Feb 26, 2012, 8:41:18 PM2/26/12
to loj...@googlegroups.com
coi

The algorithm of making a non-Lojban word into a valid Stage 3 fu'ivla
may produce homonyms.
For example, a Mexican may produce "djartako" from taco, while a
Japanese may produce "djartako" from octopus dish.
Is there algorithm to select one of them?

mi'e guskant mu'o

Pierre Abbat

unread,
Feb 26, 2012, 9:27:28 PM2/26/12
to loj...@googlegroups.com

"djartako" is a type-4, you mean "cidjrtako", but I see your point. There
isn't, and can't be, an algorithm to decide this. There can be a protocol (the
multiparty equivalent of an algorithm).

Something somewhat similar happened with Chinese provinces. There are two
called "Shanxi" in different tones. In English, one of them is arbitrarily
spelled with two a's. There's also a country with three languages called Pong.
I don't know if they have different tones, or even what kind of tones they
have.

Pierre
--
gau do li'i co'e kei do

Sid

unread,
Feb 27, 2012, 1:19:09 PM2/27/12
to loj...@googlegroups.com
ta'osai "Shaanxi" partially comes from Gwoyeu Romatzyh, a romanization
that used spelling changes instead of diacritics/numbers to indicate
tone. So, yeah.

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

guskant

unread,
Mar 2, 2012, 11:35:53 PM3/2/12
to loj...@googlegroups.com
ki'e la'oi Pierre

On Mon, Feb 27, 2012 at 11:27 AM, Pierre Abbat <ph...@phma.optus.nu> wrote:
> On Sunday, February 26, 2012 20:41:18 guskant wrote:
>> coi
>>
>> The algorithm of making a non-Lojban word into a valid Stage 3 fu'ivla
>> may produce homonyms.
>> For example, a Mexican may produce "djartako" from taco, while a
>> Japanese may produce "djartako" from octopus dish.
>> Is there algorithm to select one of them?
>
> "djartako" is a type-4, you mean "cidjrtako",

I simply used the term "Stage 3" used in the CLL 4.7:
"The rafsi categorizes or limits the meaning of the fu'ivla [...].
Such a Stage 3 borrowing is the most common kind of fu'ivla."
"Stage 4 fu'ivla do not have any rafsi classifier, and are used where
a fu'ivla has become so common or so important that it must be made as
short as possible."
I don't know another classification method.

>
> Something somewhat similar happened with Chinese provinces. There are two
> called "Shanxi" in different tones. In English, one of them is arbitrarily
> spelled with two a's. There's also a country with three languages called Pong.
> I don't know if they have different tones, or even what kind of tones they
> have.
>

Homonyms from the same language can be translated into Lojban
in distinctive fu'ivla forms with some rules. In the case of modern
Mandarin, we may create a rule that four tones [˥], [˧˥], [˨˩˦], [˥˩]
are represented by symbolic consonants l, m, n, r respectively. In the
case of violating the rule of consonant clusters, just omit the
symbolic consonant for that fu'ivla.

With this rule, two "Shanxi"s 山西 and 陕西 can be translated into Lojban
fu'ivla as tutrcanlci and tutrcanci respectively.

On the other hand, it seems that homonyms from different languages
cannot be managed with reasonable rules. However, "fu'ivla (like other
brivla) are not permitted to have more than one definition. (CLL 4.7)"

According to CLL 4.7, "Stage 3 fu'ivla can be made easily on the fly,
as lujvo can, because the procedure for forming them always guarantees
a word that cannot violate any of the rules." This statement is not
true for "djartako".

"Stage 4 fu'ivla require running tests that are not simple to
characterize or perform, and should be made only after deliberation
and by someone knowledgeable about all the considerations that apply.
(CLL 4.7)" Yes, and because of "djartako", running tests must be
applied also to Stage 3 fu'ivla.

> There
> isn't, and can't be, an algorithm to decide this. There can be a protocol (the
> multiparty equivalent of an algorithm).

Which protocol can be there?

Pierre Abbat

unread,
Mar 3, 2012, 9:40:27 PM3/3/12
to loj...@googlegroups.com
On Friday, March 02, 2012 23:35:53 guskant wrote:
> ki'e la'oi Pierre
>
> On Mon, Feb 27, 2012 at 11:27 AM, Pierre Abbat <ph...@phma.optus.nu> wrote:
> > On Sunday, February 26, 2012 20:41:18 guskant wrote:
> >> coi
> >>
> >> The algorithm of making a non-Lojban word into a valid Stage 3 fu'ivla
> >> may produce homonyms.
> >> For example, a Mexican may produce "djartako" from taco, while a
> >> Japanese may produce "djartako" from octopus dish.
> >> Is there algorithm to select one of them?
> >
> > "djartako" is a type-4, you mean "cidjrtako",
>
> I simply used the term "Stage 3" used in the CLL 4.7:
> "The rafsi categorizes or limits the meaning of the fu'ivla [...].
> Such a Stage 3 borrowing is the most common kind of fu'ivla."
> "Stage 4 fu'ivla do not have any rafsi classifier, and are used where
> a fu'ivla has become so common or so important that it must be made as
> short as possible."
> I don't know another classification method.

We decided (I'm not sure when; it may have been before I became a Lojbanist)
that a Stage 3 fu'ivla has to use a rafsi ending in a consonant. Also, if you
make a stage-4 fu'ivla but it has the form of a stage-3, you have to change
it. I've run into this once: "turndun", an Australian Aboriginal word for a
bullroarer, turned into "turndunu", but that's a type-3 word for some kind of
structure, so I changed it to "turdunu".

> On the other hand, it seems that homonyms from different languages
> cannot be managed with reasonable rules. However, "fu'ivla (like other
> brivla) are not permitted to have more than one definition. (CLL 4.7)"
>
> According to CLL 4.7, "Stage 3 fu'ivla can be made easily on the fly,
> as lujvo can, because the procedure for forming them always guarantees
> a word that cannot violate any of the rules." This statement is not
> true for "djartako".
>
> "Stage 4 fu'ivla require running tests that are not simple to
> characterize or perform, and should be made only after deliberation
> and by someone knowledgeable about all the considerations that apply.
> (CLL 4.7)" Yes, and because of "djartako", running tests must be
> applied also to Stage 3 fu'ivla.

I think you've found a bug in the Book, and it should be fixed in the next
edition. I'm not sure how to fix it, though. The bug I found in the fu'irvlazba
rules (ler(fu) + djamo (a Korean word for a letter that's a component of a
syllable) = lerndjamo by the rule as it then stood) was easy to fix (in this
case the interfix must be -l-, thus lerldjamo).

> > There
> > isn't, and can't be, an algorithm to decide this. There can be a protocol
> > (the multiparty equivalent of an algorithm).
>
> Which protocol can be there?

I don't know, but it could involve a Japanese Lojbanist talking with a Mexican
Lojbanist.

Pierre
--
li ze te'a ci vu'u ci bi'e te'a mu du
li ci su'i ze te'a mu bi'e vu'u ci

Michael Turniansky

unread,
Mar 9, 2012, 9:30:01 AM3/9/12
to loj...@googlegroups.com
Of course, it's sheer nonsense to claim that brivla can't have more
than one definition. See "facki" for example. What is really meant
is that there can't be a substantially different definition. Anyway,
there's a simple solution to the dilemma at hand. Define "djartako"
to be "x1 is filled and folded tortilla/octopus sashimi made with
additional ingredients/preparation method x2 and associated with
culture x3" There is nothing that says that fu'ivla can only have
one place, nor that it has to divide the world the way one natlang or
another does (for example, a tirxu can be a whole range of different
species).

--gejyspa

MorphemeAddict

unread,
Mar 9, 2012, 1:11:45 PM3/9/12
to loj...@googlegroups.com
What about this definition of "facki" is more than one meaning? 

x1 discovers/finds out x2 (du'u) about subject/object x3; x1 finds (fi) x3 (object)

It looks to me like several English phrases used to clarify the one meaning. 

stevo

Krzysztof Sobolewski

unread,
Mar 9, 2012, 1:43:30 PM3/9/12
to loj...@googlegroups.com
Dnia piątek, 9 marca 2012 o 19:11:45 MorphemeAddict napisał(a):
> What about this definition of "facki" is more than one meaning?
>
> x1 discovers/finds out x2 (du'u) about subject/object x3; x1 finds (fi) x3 (object)
>
> It looks to me like several English phrases used to clarify the one meaning.

Aren't:

x1 discovers/finds out x2 (du'u) about subject/object x3
x1 finds (fi) x3 (object)

two different definitions of two different concepts? To me it borders on malglico ;)
--
Ecce Jezuch
"We believe - so we're misled
We assume - so we're played
We confide - so we're deceived
We trust - so we're betrayed." - T. Haake

Jacob Errington

unread,
Mar 9, 2012, 4:51:47 PM3/9/12
to loj...@googlegroups.com
Isn't finding something equivalent to learning its location, in which case the two definitions *are* the same, where facki2 for the second defintion is {lo ka makau se zvati ce'u} ? Also, {facki} by itself seems extremely redundant to djuno, minus a tense implication ({facki} seems to me like {co'a djuno}). It also seems like {cilre} is a {co'a djuno} + built-in modal places. Same for {ctuca} (considering that the teaching succeeds, of course.)

compare:
mi facki lo du'u do jbocre [kei do]
mi facki lo du'u lo mi cukta cu zvati makau [kei lo mi cukta]
mi co'a djuno lo du'u do jbocre [kei do]
mi co'a djuno lo du'u lo mi cukta cu zvati makau [kei lo mi cukta]
mi cilre lo du'u do jbocre [kei do la irk lo nu mi'o casnu bau lo lojbo]
do ctuca mi lo du'u do jbocre [kei do lo nu mio casnu bau lo lojbo]

Now, it seems like the difference between {cilre} and {ctuca} is volition on the part of ctuca1/cilre4.
Here's my lojban definition of what {ctuca} means:
lo ka ce'u goi ko'a ctuca ce'u goi ko'e ce'u goi ko'i ce'u goi ko'o ce'u goi ko'u cu ka ce'u goi ko'a zukte ce'u goi ko'u lo nu ce'u goi ko'e cilre ce'u goi ko'i ce'u goi ko'o ko'u

IMHO determining the location of an object is more akin to losing that object than "beginning to know something". Thus {cirko} and {to'e cirko}/{tolcri} are more appropriate.

mi cirko tu'a lo mi cukta
mi tolcri tu'a lo mi cukta

(I'm working under the anal-retentive assumption that cirko is "lose property" and tolcri is "acquire property" (hence the sumti-raising), in which case in a bare-bones lojban, {tolcri} is equivalent to {binxo})

mu'o mi'e la tsani

2012/3/9 Krzysztof Sobolewski <jez...@interia.pl>

djandus

unread,
Mar 10, 2012, 10:56:00 AM3/10/12
to loj...@googlegroups.com
> "Stage 4 fu'ivla require running tests that are not simple to
> characterize or perform, and should be made only after deliberation
> and by somepony knowledgeable about all the considerations that apply.

> (CLL 4.7)" Yes, and because of "djartako", running tests must be
> applied also to Stage 3 fu'ivla.

I think you've found a bug in the Book, and it should be fixed in the next 
edition.

hmm... do you (ro do) think this is a bug in that the CLL should not say that, or that the system for Stage 3 is bugged?
Personally, I see this like so:
  1. Person M loves tacos. M defines {djartako} to refer to tacos. M uses {djartako} with all his friends.
  2. Person J loves octopus shashimi. J defines {djartako} to refer to octupus.
  3. J notices clash, tells M
  4. (The Policy) They discuss things and hammer out a solution, one of:
    • Our definitions are similar enough, let's combine them into the same word. (see gejyspa)
    • Our definitions are wildly different. Let's add optional rafsi that clarify the meaning either by meaning or by cultural heritage. (Optional in the sense that you may use {djartako} when context makes the meaning clear.)
    • Our definitions disagree, and we really have a long word already. (not the case here) Let's arbitrarily make up a distinction, maybe by changing a random character. (I thoroughly dislike this option)
  5. Much rejoicing.
This is just how I see options for the policy could run. Does anyone have additions?

Also, back to fixing the CLL, correct me if I'm wrong, but either we simply need to alter the originally quoted error and add text about a policy, or we decide that the system for constructing Stage 3 is at fault (unlikely) and go and do heavy rewrites. IMHO, I don't see why this would warrant the latter, huge rewrites.

.i ta'o

 Of course, it's sheer nonsense to claim that brivla can't have more
than one definition.  See "facki" for example.

How I've always chosen to define the claim that brivla can't have more than one definition is something like: "The space of meanings for a brivla must be connected, continuous, and smooth."
Though clearly more of a mathy definition than linguistic, it serves my purposes well. By "space" I imagine a set-like object sitting inside the space of all possible meanings, where different meanings have varying likelihoods of being correctly described with the word being defined. "Continuous" and "smooth" are just ruling out blatantly terrible definition structures, things like "You may use {facki} for any type of finding, except for finding Russians. For that, use <arbitrary_word_here>." I imagine the space of this terrible facki definition would be smooth, continuous, and pretty, excepting a blatant hole violently ripped out of the middle of it. The important part, relevant to our discussion here, in "connected." By this, I mean that if a meaning must be clarified with multiple descriptions, these descriptions are only allowed to either narrow down the meaning or broaden it, not add a separate one. In this idea, for instance, the definition of {facki} is perfectly fine:
x1 discovers/finds out x2 (du'u) about subject/object x3
x1 finds (fi) x3 (object)
The second statement serves to offer a syntactical benefit by providing a default x2 (when providing an x3) that forces this particular meaning. The meaning provided by the second statement is entirely contained within the far more general statement prior, (as demonstrated by tsani with {lo ka makau se zvati ce'u},) and so it's not providing an unconnected meaning, just a helpful shorthand for a special case. Given, the choice for the default x2 is borderline malglico IMO, but I also think it doesn't "provide more than one definition" or is particularly terrible.
Compare this to, say, the definition of cramp. Like many English words, it has multiple part of speech meanings that have been spreading apart over the years, and even within one part of speech, there are definitions that actually are disconnected. Sorry for the cultural necessity here, but English speakers know that the three noun definitions:
1: a painful involuntary spasmodic contraction of a muscle
: a temporary paralysis of muscles from overuse — compare writer's cramp
3
a : sharp abdominal pain —usually used in pluralb : persistent and often intense though dull lower abdominal pain associated with dysmenorrhea —usually used in plural
are each actually separate things, not different subexamples. Here's a few checks to prove it:

A) Which definition above is the most general? (1) looks good, but it doesn't involve the paralysis necessary in (2). There is no general definition, just a few (arguably two) disconnected meaning spaces. 

B) Imagine a case where you use each definition. This isn't like a sentence or a conversation, but the situation. For instance, for me, I think of (1) waking up cramping in the middle of the night (2) writing in class and my hoof cramps (3) having very specific symptoms that I don't have much experience with. Now, imagine that each situation had a separate word, and you tried to use the wrong word to describe the situation. For instance, if I said I woke up in the middle of the night with my leg cramping, but I used the word for (2), then someone listening would wonder how on earth I overused my leg while I was asleep. Thus, (1) is not a subset of (2). If I described my hand cramping using (1), there'd be a distinct lack of spasming. Thus, (2) is not a subset of (1). (3), I don't really know about.

The point of all of this is that Lojban definitions of words shouldn't have this trouble at all. You should be able to interpret a Lojban definition with one general idea, with maybe a few narrowing or broadening specifications, not a collection of separate examples. Which is why I actually read/interpret Lojban definitions differently than English ones. I read Lojban defs trying to keep one idea at stake and using additional information only to mold that one idea; I read English ones expecting each def to provide at least a semi-unique usage, which I append to a list of ideas attached to that word.

.i ta'onai
Sorry about that long tangent!

djandus

.arpis.

unread,
Mar 10, 2012, 1:04:21 PM3/10/12
to loj...@googlegroups.com
"The space of meanings for a brivla must be connected, continuous, and smooth."
.u'i .ie sai

> --
> You received this message because you are subscribed to the Google Groups
> "lojban" group.

> To view this discussion on the web visit
> https://groups.google.com/d/msg/lojban/-/NPDNzpRr6uUJ.


>
> To post to this group, send email to loj...@googlegroups.com.
> To unsubscribe from this group, send email to
> lojban+un...@googlegroups.com.
> For more options, visit this group at
> http://groups.google.com/group/lojban?hl=en.

--
mu'o mi'e .arpis.

MorphemeAddict

unread,
Mar 10, 2012, 1:23:44 PM3/10/12
to loj...@googlegroups.com
+1 stevo

Michael Turniansky

unread,
Mar 19, 2012, 8:46:15 AM3/19/12
to loj...@googlegroups.com
  +2  Although the contention that the default selfacki when missing is a particular one that English assigns to the word that glosses facki is, perforce, malglico (not stopping me from using zo facki in that manner, I most certainly do.  But I must agree that it's definitely not a culturally neutral default).

              --gejyspa
Reply all
Reply to author
Forward
0 new messages