Are Natlang the best case for entropy in communication ?

[Escape Landsome]

Coi rodo,

in a noisy verbal interaction, for instance, on a phone call (but this
applies to any case, indeed), it is good for communication that the
elements of signal can be easily distinguished, so as to avoid
reception errors.

This is achieved by natlangs, by sparsing all the existing words (for
instance, adverbs) in such a "morphologic space" that has very few
collisions. For instance, the english words for "few", "many", "a
lot" and "none" are phonologically very different of each other, so
there is little chance you could confuse them by hearing them on a
deficient phone.

But this is not the case of lojban words, for instance so'a, so'e,
so'i, so'o etc. are very near of each other, and, assuming you don't
hear well the last vowel, you could infer something very far from what
was intended by the other speaker.

So, is not that something that is annoying ?

[la .lindar.]

So, is not that something that is annoying ?

No.

It is quite annoying.

What is your point, though? 

[John E Clifford]

Presumably the oft-repeated one to take noisy signals into account in LoCCan 3 so that very different concepts don't have readily confused words (a problem across most cmavo categories, at least).

---

From: la .lindar. <lindar...@gmail.com>
To: loj...@googlegroups.com
Sent: Monday, June 18, 2012 12:28 PM
Subject: [lojban] Re: Are Natlang the best case for entropy in communication ?

--
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/-/KIe4XBEjVE8J.
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.

[John E Clifford]

Or, maybe, that Lojban sucks because it didn't (despite warnings).

---

From: John E Clifford <kali9...@yahoo.com>
To: "loj...@googlegroups.com" <loj...@googlegroups.com>
Sent: Monday, June 18, 2012 1:18 PM
Subject: Re: [lojban] Re: Are Natlang the best case for entropy in communication ?

[Escape Landsome]

Escaping sarcastically the difficulty does not solve it either.

I understand that lojbanists use to protect themselves against any
criticism, as in a besieged citadel, but the point still holds : <<
if very different concepts are given very near phonological forms,
isn't this a bad move ? >>

And : << Does not the fact natlangs do not have this problem generally
speaking imply that they are more well designed than lojban on this
particular point ? >>

[Jonathan Jones]

On Tue, Jun 19, 2012 at 2:02 AM, Escape Landsome <esca...@gmail.com> wrote:

Escaping sarcastically the difficulty does not solve it either.

Who's being sarcastic?
 

I understand that lojbanists use to protect themselves against any
criticism, as in a besieged citadel, but the point still holds :  <<
if very different concepts are given very near phonological forms,
isn't this a bad move ? >>

I can think of no case where "very different concepts are given very near phonological forms". It is true that there are many cmavo which are very nearly the same, such as FA, SE, etc., but all of these groupings are very closely related to each other, differing in very minor ways.

Your own example of the so'V series is not nearly as "different" as you seem to think. Each of them is a point on the scale All-None, exclusive. The only difference is where on that scale each is, with so'a being closest to All, and so'u closest to None.
 

And : << Does not the fact natlangs do not have this problem generally
speaking imply that they are more well designed than lojban on this
particular point ? >>

I don't think so, mainly because, being NATlangs, they weren't designed at all. I highly doubt that the phonological properties of words were ever taken into account during the etymological evolution of those languages. In fact, there is ample evidence to the contrary, as there are many cases where words that are /extremely/ different concepts have /extremely/ similar phonotactics, such as the English to, too, and two. I would argue that mistaking, for example, "bow" (either the act-of or the ship-part, but not the clothing) and "now", or "to" and "too" is much more damaging to listener comprehension than mistaking "so'a" for "so'u".

--
mu'o mi'e .aionys.

.i.e'ucai ko cmima lo pilno be denpa bu .i doi.luk. mi patfu do zo'o
(Come to the Dot Side! Luke, I am your father. :D )

[Escape Landsome]

> I can think of no case where "very different concepts are given very near
> phonological forms". It is true that there are many cmavo which are very
> nearly the same, such as FA, SE, etc., but all of these groupings are very
> closely related to each other, differing in very minor ways.
>
> Your own example of the so'V series is not nearly as "different" as you seem
> to think. Each of them is a point on the scale All-None, exclusive. The only
> difference is where on that scale each is, with so'a being closest to All,
> and so'u closest to None.

The paradigm is the same (all so'V are related to scale All-None) but
inside this paradigmatic choice, they are very distinct.

All is not at all the same thing as None or as Few, and if somebody is
not well heard when saying the word, the consequences are great !

As if I said [ x = 0% ], [ x = 10% ], [ x = 100% ], and all the onus
of communication lied on the real variable near the percent sign...
Communication would be in great risk to be lost.

In the case of natlang some redundancy is set to avoid this, namely
the "f-" or "few" opposes the "n-" of "none", but also "-ew" of "few"
opposes "-one" of "none", so that, if ever one phonem is not well
understood, the other ones are there to save the day.

>> And : << Does not the fact natlangs do not have this problem generally
>> speaking imply that they are more well designed than lojban on this
>> particular point ? >>
>
>
> I don't think so, mainly because, being NATlangs, they weren't designed at
> all. I highly doubt that the phonological properties of words were ever
> taken into account during the etymological evolution of those languages. In
> fact, there is ample evidence to the contrary, as there are many cases where
> words that are /extremely/ different concepts have /extremely/ similar
> phonotactics, such as the English to, too, and two. I would argue that
> mistaking, for example, "bow" (either the act-of or the ship-part, but not
> the clothing) and "now", or "to" and "too" is much more damaging to listener
> comprehension than mistaking "so'a" for "so'u".

Natlangs were designed, but the designer is a non-human (and
non-divine) random process of natural selection. Natural selection
favorises random creations, but random creations naturally occupy the
phonological space smoothly and in a sparse way, so one can say,
natural selection naturally designs words that are good for efficient
communication.

[Jonathan Jones]

On Tue, Jun 19, 2012 at 2:53 AM, Escape Landsome <esca...@gmail.com> wrote:

> I can think of no case where "very different concepts are given very near
> phonological forms". It is true that there are many cmavo which are very
> nearly the same, such as FA, SE, etc., but all of these groupings are very
> closely related to each other, differing in very minor ways.
>
> Your own example of the so'V series is not nearly as "different" as you seem
> to think. Each of them is a point on the scale All-None, exclusive. The only
> difference is where on that scale each is, with so'a being closest to All,
> and so'u closest to None.

The paradigm is the same (all so'V are related to scale All-None) but
inside this paradigmatic choice, they are very distinct.

All is not at all the same thing as None or as Few, and if somebody is
not well heard when saying the word, the consequences are great !

As if I said [ x = 0% ], [ x = 10% ], [ x = 100% ], and all the onus
of communication lied on the real variable near the percent sign...
Communication would be in great risk to be lost.

In the case of natlang some redundancy is set to avoid this, namely
the "f-" or "few" opposes the "n-" of "none", but also "-ew" of "few"
opposes "-one" of "none", so that, if ever one phonem is not well
understood, the other ones are there to save the day.

Except that is not the case.

Which of the following cases of listener mishearing would you consider to be of greater consequence?

"I have two sheep because I'm tired."

{lonu lo citri cu na morji piso'uroi cu gasnu lonu lo citri cu rapli}

>> And : << Does not the fact natlangs do not have this problem generally
>> speaking imply that they are more well designed than lojban on this
>> particular point ? >>
>
>
> I don't think so, mainly because, being NATlangs, they weren't designed at
> all. I highly doubt that the phonological properties of words were ever
> taken into account during the etymological evolution of those languages. In
> fact, there is ample evidence to the contrary, as there are many cases where
> words that are /extremely/ different concepts have /extremely/ similar
> phonotactics, such as the English to, too, and two. I would argue that
> mistaking, for example, "bow" (either the act-of or the ship-part, but not
> the clothing) and "now", or "to" and "too" is much more damaging to listener
> comprehension than mistaking "so'a" for "so'u".

Natlangs were designed, but the designer is a non-human (and
non-divine) random process of natural selection.   Natural selection
favorises random creations, but random creations naturally occupy the
phonological space smoothly and in a sparse way, so one can say,
natural selection naturally designs words that are good for efficient
communication.

Based entirely on my own experience with English, I would have to disagree. I do not consider English at all suited for "efficient communication".

[Jonathan Jones]

On Tue, Jun 19, 2012 at 3:10 AM, Jonathan Jones <eye...@gmail.com> wrote:

Except that is not the case.

Which of the following cases of listener mishearing would you consider to be of greater consequence?

"I have two sheep because I'm tired."

{lonu lo citri cu na morji piso'uroi cu gasnu lonu lo citri cu rapli}

Keep in mind both of these examples differ from intention by exactly one phoneme.

[Escape Landsome]

The problem is not to find examples of utterances where one single
phonem is harmful

You WILL find some

There are SOME

In any language

You could even say, for instance, that "a few" is too much near "few"

But the problem is that here you have a whole set of possibilities (=
a paradigm) where all the possibilities are very near one to another,
and yet their meanings are very distinct and even, opposite.

I don't find an hypothetic language that would have numbers 1, 2, 3,
... 1000 and which would be gah, gahh, g'ah, gaa, gââ could be very
cleverly designed, especially if the other words are very different
form "gah"

I don't find lojban is greatly inspired to have "none", "all", and
"many" and "few" distinct by only one letter, in the same paradigm !

[Escape Landsome]

Please keep in mind a great deal of the difficulty comes form the fact
that the words that can be confounded come in the same paradigm, that
is "none" versus "all" for instance

It would not be as bad if it were "sheep" versus "two", or "spinach"
versus "to sleep"

[gleki]

On Monday, June 18, 2012 1:52:59 PM UTC+4, Escape Landsome wrote:

Coi rodo,

in a noisy verbal interaction, for instance, on a phone call (but this
applies to any case, indeed), it is good for communication that the
elements of signal can be easily distinguished, so as to avoid
reception errors.

This is achieved by natlangs, by sparsing all the existing words (for
instance, adverbs) in such a "morphologic space" that has very few
collisions.   For instance, the english words for "few", "many", "a
lot" and "none" are phonologically very different of each other, so
there is little chance you could confuse them by hearing them on a
deficient phone.

Yes, but here comes another problem. Memorising  "so'V" takes less time than  "few", "many", "a 

lot" and "none" cuz they are ... phonologically very different of each other!

But this is not the case of lojban words, for instance so'a, so'e,
so'i, so'o etc. are very near of each other, and, assuming you don't
hear well the last vowel, you could infer something very far from what
was intended by the other speaker.

I believe that most Chinese words sound the same for an ordinary English speaker because the latter is just not used to it's phonology.

But still Chinese is the language with which you can send people to outer space. Does it mean that this language is bad?
Let's get rid of such subtle differences in LoCCan 3.

But then other people will appear and say that <l> and <r> sound the same for them.

<m> and <n> sound the same. In LoCCan ~10.0 we'll have <p,t,k,m,a,i,u> sounds  only.

Is it what we want?

Why not try pronouncing {so'i} and  {so'e} more carefully than in English?

Chinese phonology requires "singing" sounds because of tones (at least that's how many Europeans perceive it).

So why not doing the same in Lojban?

Am I not right about LoCCan 3? May be all cmavo should have the form CVCCV/CCVCV?

[Escape Landsome]

> I believe that most Chinese words sound the same for an ordinary English
> speaker because the latter is just not used to it's phonology.
> But still Chinese is the language with which you can send people to outer
> space. Does it mean that this language is bad?

No, you miss the point.

The problem is not with pronounciation in itself but with *phonemes*.

Phonemes are defined as smallest units of phonological type that carry
meaning in a double segmentation scheme.

That is, /b/ and /p/ are english phonemes, not because they're not the
same sounds, but because there is a pair of words such as "bit" /
"pit", and thus it is wise to consider an opposition between them.

But in other languages, such as chinese, b and p are not distinct phonemes.

Which is no probleme, because Chinese has other meaningful opposition
pairs, such as "plosive" vs "mute", or "nasalized" versus "no-nasal"

.

And thus, /a/, /e/, /i/, /o/ /u/ are phonemes in lojban in the
contradistinction set of so'V

But different words with opposite meanings IN THE SAME PARADIGME
differ form only one phoneme, which is not wise

---

(Also, "a lot", "few", "all" and "none" do not differ only from one
feature in Chinese... but from a lot a differente phonological
feature. We don't care if "all" and "horse" are very near. But we
care this be the case for "all" and "none" !!!)

[John E. Clifford]

Well, a lot of "words" in Chinese are phonologically identical, but the problem is solved by not using a given word in isolation but rather embedded in a cloud of distinctive other words. When a Chinese speaks of shrimp, for example, the expression he uses translates as something like "shrimp shrimp [a different word] fish bug" (from memory, details may vary, though not the principle). The point is that languages have various ways around unclarity but it is not obvious what Lojban's are in the cmavo cases, at least. This matter is scarcely unique to Lojban, even to the point of having been warned early on: toki pona for "all" is 'ale' and for "no" is 'ala', clearly worse than 'ro' and 'no'.


Sent from my iPad

[Escape Landsome]

> Well, a lot of "words" in Chinese are phonologically identical, but the problem is solved by not using a given word in isolation but rather embedded in a cloud of distinctive other words.  When a Chinese speaks of shrimp, for example, the expression he uses translates as something like "shrimp shrimp [a different word] fish bug" (from memory, details may vary, though not the principle).

True, that is, with redundancy the code can convey information more
securely... This is well known in information theory.

Please, consider the notion of paradigm. A paradigm is a set of
options or choices you can choose one within the set. If there is
ambiguity between two words that are not in the same paradigm, this is
not as annoying as if the ambiguit lies between two words of the same
paradigm, not even mentioning the case of two opposite words !

For instance, in english, you could confound "none" and "nun", or
"all" and "owl", but at least there's no chance to confound "all" and
"none", AND THAT IS THE IMPORTANT THING !

[Jonathan Jones]

I contend that confusing words of separate "paradigms" is worse.

Moreover, I contend that you are making a mountain out a molehill, as such problems as you are calling attention to are both rare and easily fixed.

The pronunciation of Lojban's vowels were specifically chosen to be as unlike in sound to the others as possible. .abu sounds nothing like .obu, .ebu sounds nothing like .ibu, and so on. In light of this, I submit that the only case where confusion between which of a series is the one that was said is even likely is when the distincting vowel is completely unheard, and in such a case as that, either the listener would either attempt to make an educated guess as to which of the series was most likely to have been the spoken one based upon the context of the utterance it was a part of, or alternatively asking the speaker to repeat that particular word.

If you do not understand a person, it is a simple matter to ask for clarification.

[Escape Landsome]

> I contend that confusing words of separate "paradigms" is worse.

And I contend that.

If the confusion arises between two words which pertain to different
paradigm for instance :

QUANTITY = { all, some, many... } and BIRD = { owl, pigeon, crow... }

then it is easy to disambiguate "all" from "owl" because the "???"
(which is = all or = owl) is a quantity or a bird in context, but most
rarely both.

But when the two confusionable words come from the same paradigm-set

QUANTITY = { so'a, so'e, so'i... }

the confusion is VERY harmful

[la .lindar.]

I think what  I was really trying to get down to is this:

So?

What's your point?

[Escape Landsome]

> So?
> What's your point?

This comes as a direct consequence of my question, don't you see it ?

This makes us wonder if words should not be created in some way that
allows more entropy.

[Jonathan Jones]

I have to correct you on something you said:

On Tue, Jun 19, 2012 at 11:22 PM, Escape Landsome <esca...@gmail.com> wrote:

This makes me wonder if words should not be created in some way that
allows more entropy.

 

[gleki]

On Tuesday, June 19, 2012 2:58:52 PM UTC+4, Escape Landsome wrote:

> I believe that most Chinese words sound the same for an ordinary English
> speaker because the latter is just not used to it's phonology.
> But still Chinese is the language with which you can send people to outer
> space. Does it mean that this language is bad?


No, you miss the point.

The problem is not with pronounciation in itself but with *phonemes*.

Phonemes are defined as smallest units of phonological type that carry
meaning in a double segmentation scheme.

That is, /b/ and /p/ are english phonemes, not because they're not the
same sounds, but because there is a pair of words such as "bit" /
"pit", and thus it is wise to consider an opposition between them.

But in other languages, such as chinese, b and p are not distinct phonemes.

Which is no probleme, because Chinese has other meaningful opposition
pairs, such as "plosive" vs "mute", or "nasalized" versus "no-nasal"

.

And thus, /a/, /e/, /i/, /o/ /u/ are phonemes in lojban in the
contradistinction set of so'V

But different words with opposite meanings IN THE SAME PARADIGME
differ form only one phoneme, which is not wise

Who is right? The one who says that words must be easy to  memorise?

Or the one who says that there should be some entropy?

No language can be convenient for everyone.

[tijlan]

Below are my notes I've been putting down that I'm posting just now.
Don't bother to respond to where it overlaps with other jbopre's
points.

On 18 June 2012 10:52, Escape Landsome <esca...@gmail.com> wrote:
> This is achieved by natlangs, by sparsing all the existing words (for instance, adverbs) in such a "morphologic space" that has very few collisions.

Also in a syntactic space. "To get hold of X", for example, can be
seen as a syntactically diffused equivalent of "to take X" (the latter
is sonically shorter and can therefore be harder to pick up in a bad
speech environment).

> But this is not the case of lojban words, for instance so'a, so'e,
> so'i, so'o etc. are very near of each other, and, assuming you don't
> hear well the last vowel, you could infer something very far from what
> was intended by the other speaker.
>
> So, is not that something that is annoying ?

My practice of Lojban is mostly textual, and the sonic disadvantage in
question (if any) may have been offset by the words' mnemonic
advantage for me (i.e. the set of {so'a, so'e, so'i, so'o, so'u} would
be easier to memorize than the set of corresponding natlang words with
incoherent dissimilar shapes).

The sememe-per-phoneme (information-per-sound) ratio, or information
density, can be as much relevant to the communicability of spoken
words. It would seem: the higher this ratio is, the more information
listeners are likely to miss out from a speech in a noisy place or on
a bad phone line. Lojban seems to have a high ratio, with its vocally
minimal lexicon. But that isn't to say natlang utterances are
altogether equally easier to scoop out of noise than Lojban's are.
Here's an interesting study:

http://www.time.com/time/health/article/0,8599,2091477,00.html

information density | average spoken syllables per second
.94 | 5.18 -- Mandarin
.91 | 6.19 -- English
.63 | 7.82 -- Spanish
.49 | 7.84 -- Japanese

A higher information density appears to be countered by more-stretched
syllabism (more diphthongs & consonants in the case of English). But
does that help in a noisy situation? As a non-native English speaker
living in UK, I often have a hard time figuring out what people are
saying on the phone, on top of the fact that they come off as speaking
rather fast overall in spite of the supposedly stretched syllables.
Vocal elements are largely idiomatized, sometimes even gestalt-ish,
requiring listeners to make up or convert in their mind the elided or
altered sounds:

[haɪʔ pɑːʔ] --> Hyde Park
[bəiʔ] --> beat
[bɒvə] --> bother
...

This natlang phenomenon... Is it entropic? Is it convenient?

On 19 June 2012 09:02, Escape Landsome <esca...@gmail.com> wrote:
> I understand that lojbanists use to protect themselves against any
> criticism, as in a besieged citadel, but the point still holds : <<
> if very different concepts are given very near phonological forms, isn't this a bad move ?

{so'a, so'e...}, classified as PA4, conceptually or functionally
aren't so miscellaneous. The same for {ko'a, ko'e...}, {fa, fe...},
{la, le...}, {noi, poi...} etc.

A group of "very different concepts", for me, would be something like:

many
gentle
monitor speaker
week
logical fallacy
hat-trick
central nervous system
...

> And : << Does not the fact natlangs do not have this problem generally
> speaking imply that they are more well designed than lojban on this
> particular point ? >>

I don't know that's a fact. What I know is that the following can
confuse foreign English learners:

crush, crash, clash
run, ran
hit, hit
decision, discission
complement, compliment
site, sight
discrete, discreet
flare, flair
vice-president, a life of vice
...

On 19 June 2012 09:53, Escape Landsome <esca...@gmail.com> wrote:
> Natlangs were designed, but the designer is a non-human (and
> non-divine) random process of natural selection.

Language use is largely filtered by human intelligence. Words to be
used (and recorded) can be selected non-randomly, even altered if
deemed necessary or effective for some purpose:

shit --> shite
fucking --> effing
all correct --> ok
laughing out loud --> lol

After all, modern English owes a lot to artistic writers such as
Shakespeare, for whom a linguistic expression could well be an
engineering process.

Many 'natlangs' have official regulators. French orthography, for
instance, has been occasionally reformed to improve both textual and
sonic consistencies according to the lexical families.

> Natural selection
> favorises random creations, but random creations naturally occupy the
> phonological space smoothly and in a sparse way, so one can say,
> natural selection naturally designs words that are good for efficient
> communication.

What's efficient is determined by the environment. Apes' hands are
efficient in forests but not in seas. Our linguistic environment has
been massively changing through the inventions of printing, internet,
etc., and we can question how well are natlangs keeping up with this
shift in media ecology at such an increasing rate.

Also crucial is the changing social structures & norms. How often do
we say "he or she" for the lack of a better pronoun to refer to the
increasingly common gender-irrelevant sets of individuals? Would you
expect a 'random natural' process to give birth to a good neuter to
replace this inefficient three-word phrase?


mu'o

[Jonathan Jones]

On Wed, Jun 20, 2012 at 4:13 AM, tijlan <jbot...@gmail.com> wrote:

Also crucial is the changing social structures & norms. How often do
we say "he or she" for the lack of a better pronoun to refer to the
increasingly common gender-irrelevant sets of individuals? Would you
expect a 'random natural' process to give birth to a good neuter to
replace this inefficient three-word phrase?

Since that process is causing the replacement to be "they", I personally say no.

[John E. Clifford]

Well, as a suggestion against the distant day that Lojban might be used in a noisier channel than print and small group hyper focused conversation, someone might think a bit about what redundancy devices (if any) Lojban has for potentially confusing slots.  Since this day is unlikely to occur, this is merely a theoretical exercise, but at least we could say we have thought of the matter the next time issue comes up (less often than once every five years, I think, though no one seems to have been moved to action by the last couple of rounds either).

Sent from my iPad

On Jun 20, 2012, at 12:12 PM, "la .lindar." <lindar...@gmail.com> wrote:

Maybe I'm not being super to-the-point here, so let me be even more rude and blunt:

Stop being an asshole. We all know that Lojban has problems. You're harping on this like you expect everybody to drop everything and say, "OH FUCK! Well, I guess you're right. God damn, now we have to start over...". I'm asking you what your point is because I can only extrapolate that you expect everybody to give up on Lojban and start from scratch keeping your point in mind. If that isn't your point, then what the fuck are you getting at; what do you expect us to do with this information? If it is your point, then we all get it (and knew that thousands of years ago) and you can stop bleeding the wound.

--
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/-/jj37asHsvYcJ.

[Escape Landsome]

> Stop being an asshole. We all know that Lojban has problems. You're harping

> on this like you expect everybody to drop everything and say, [SNIP]

> and you can stop bleeding the wound.

You don't need to be blunt.

You react like a believer whose faith is attacked.

In any case, I don't attack YOU, nor even your "faith", but would you
say that a man that would complain to Russell about problems with the
set y = {x | x ∉ x} would be "an asshole" ?

Well, he would be if you take the things too much emotionally, but I
rather think he would be a great scientist.

Here I point the necessity of adapting the mechanism of redundancy in
lojbanic communication. There IS some problem. The fact that
Natlangs do have some glimpses of this problem too is not relevant,
because in most case this concerns pairs of concepts very near such as
"ran"/"run" where the unguessable feature is [PAST]/[PRESENT]. But
here so'V = QUANTITY = { all, none, many, a few ... } is A WHOLE
PARADIGM.

It's up to you to consider I overlook that, but I tend to think you
are too much emotionally involved in your reaction. Ought to be more
scientifically-minded...

[gleki]

On Thursday, June 21, 2012 11:01:13 AM UTC+4, Escape Landsome wrote:

> Stop being an asshole. We all know that Lojban has problems. You're harping
> on this like you expect everybody to drop everything and say, [SNIP]
> and you can stop bleeding the wound.

You don't need to be blunt.

You react like a believer whose faith is attacked.

In any case, I don't attack YOU, nor even your "faith", but would you
say that a man that would complain to Russell about problems with the
set y = {x | x ∉ x} would be "an asshole" ?

Well, he would be if you take the things too much emotionally, but I
rather think he would be a great scientist.

Here I point the necessity of adapting the mechanism of redundancy in
lojbanic communication.   There IS some problem.   The fact that
Natlangs do have some glimpses of this problem too is not relevant,
because in most case this concerns pairs of concepts very near such as
"ran"/"run" where the unguessable feature is [PAST]/[PRESENT].   But
here so'V = QUANTITY = { all, none, many, a few ... } is A WHOLE
PARADIGM.

So what is your suggestion?

Should we perform a scrutiny of  all cmavo and think how each selmaho should sound like?

What about gismu?

What about lujvo (where low signal-to-noise ratio may also be important)?

May be you already have something specific in mind?

[tijlan]

On 21 June 2012 08:01, Escape Landsome <esca...@gmail.com> wrote:
> In any case, I don't attack YOU, nor even your "faith", but would you
> say that a man that would complain to Russell about problems with the
> set y = {x | x ∉ x} would be "an asshole" ?
>
> Well, he would be if you take the things too much emotionally, but I
> rather think he would be a great scientist.

Except that the aspect you are characterizing as a 'problem' of
Lojban, is conditional. It could be a problem in particular (e.g.
noisy) situations but not in other situations. If we are to try to be
good scientists, we have to first determine whether there is a reason
to believe that the conditions for the suspected problem are or will
be common enough to warrant a remodelling or disposal of the subject.
Otherwise it's like blaming aircraft for having the possibility of
stall at an unnecessary altitude.

> The fact that
> Natlangs do have some glimpses of this problem too is not relevant,
> because in most case this concerns pairs of concepts very near such as
> "ran"/"run" where the unguessable feature is [PAST]/[PRESENT].   But
> here so'V = QUANTITY = { all, none, many, a few ... } is A WHOLE
> PARADIGM.

"Run/ran" as well as "push/pushed" etc. belong to the paradigm of
TENSE, just as {so'a/so'e/...} belong to the paradigm of QUANTITY. For
each group, the meaning can be put at stake by the lack of one
particular phoneme.

In the cases of
cost/cost
cut/cut
hit/hit
hurt/hurt
let/let
put/put
quit/quit
set/set
shut/shut
...
, no phonemic clue is available at all.


mu'o

[Escape Landsome]

We have no reason to get upset because on ONE PARTICULAR point
natlangs behave better than lojban.

There is even a chance that lojban can be amended in a way it behaves
better then natlangs after the amendment.

Suppose for instance we are given a new consonant "q", one could state
that from now there is a strict equivalence between the phonological
sequences

"a" and "aqa"

"e" and "eqe"

"i" and "iqi"

"o" and "oqo"

"u" and "uqu"

Thus, so'a and so'e can be confused in a noisy environement, but
saying so'aqa ou so'eqe would avoid that.

This is a simple example to show you that discussing a drawback of
lojban does not mean being mean towards it, rather it is what is
expected from anybody here : that is, being scientific and examine
closely and open-mindedly any problem.

[I don't think the solution I proposed is a good one, either. But at
least it shows this is no dead-end street question. And also we've
no need to be aggressive. Meanwhile, natlangs are still better than
lojban on the entropy topic]

[Muhammad Na'el]

In the simplest possible manner, I like this; can't remember seeing it before...

--
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.

--

›› MN

[gleki]

On Thursday, June 21, 2012 2:49:12 PM UTC+4, Escape Landsome wrote:

We have no reason to get upset because on ONE PARTICULAR point
natlangs behave better than lojban.

There is even a chance that lojban can be amended in a way it behaves
better then natlangs after the amendment.

Suppose for instance we are given a new consonant "q", one could state
that from now there is a strict equivalence between the phonological
sequences

"a" and "aqa"

"e" and "eqe"

"i" and "iqi"

"o" and "oqo"

"u" and "uqu"

Oh. If you replace VqV with V3 (in Mandarine third tone, where V - vowel) that would be quite reasonable too.

[gleki]

And forgot to add that this suggestion just changes phonology of certain phonemes if we consider V=VqV as one phoneme no matter what sound {q} denotes.

This can be reasonable if we take into account that many lerfu have several possible phonemic realisations each (most notably "th/h" for { ' }).

[John E. Clifford]

The "natural" response to a problem of the simple sort imagined here is repetition with added stress: 'so'E', which falls within this scheme.  There are harder cases where this is not enough (and even this breaks the strict written-spoken isomorphy).  But looking for fixes is a reasonable approach, given that prevention is no longer an option.  And, of course, the problem is still pretty remote.

Sent from my iPad

--

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/-/CRtH6Y33AxwJ.

[Jonathan Jones]

> It's up to you to consider I overlook that, but I tend to think you
> are too much emotionally involved in your reaction.   Ought to be more
> scientifically-minded...

I have to laugh at this. For one thing, Lindar isn't reacting to your argument. As he said, he and every one else KNOWS what your argument is. His response was "What's your point?", which means he wants to know WHY you're "harping" on this. His anger, if any, is at the fact that he has asked you repeatedly, and you have yet to answer him, but instead continue to repeat yourself.

Furthermore, YOU are not being scientific.

The scientific method is:
1)  Formulate a question. In this case, I believe the question is "What are the cause(s), if any, that make Lojban more prone to listener error in noisy environments?"

2) Make a hypothesis. "The high similarity of the various cmavo series, such as for example [so'a, so'e, so'i, so'o, so'u], make Lojban more error prone."

3) Create predictions from the hypothesis. "A person hearing {lonu lo citri cu na morji piso'aroi cu gasnu lonu lo citri cu rapli} in a noisy environment will think he heard any of [so'a, so'e, so'i, so'o, so'u], but will be unable to be sure which of them it was."

4) Perform tests.

5) Analyze the test results, adjusting the hypothesis accordingly and returning to step 3).

You stopped at 2) and said "BIG PROBLEM! EVERYONE LISTEN TO ME NOW!" Not only do you simply assume you're correct without even attempting to verify your claim, but you offer nothing to fix this "problem" as well.

I don't believe it's a problem. Even assuming it is a problem, I don't believe it's nearly as important or large a problem as you seem to think, which is why I previously accused you of making mountains of molehills. Finally, even assuming that you are completely correct and this is a huge extremely important issue,  you offer no solution to the problem, nor do you even state why you've brought it up /in the first place/.

Now, we all know that no changes that are not *absolutely* necessary will even be considered at this time, or in the foreseeable future, so obviously nothing's going to be done about this issue, assuming it exists. I've already expressed my opinion to the contrary.

However, even were we considering proposals for changes, this one would not *ever* pass. The reasons for this are: any change would break current and past Lojban text (not backwards compatible), it has not been shown that this issue actually exists, and it is a matter for debate whether the unproven issues are more important than the well-known, established benefits (i.e., ease of learning).

In any case, as far as I'm concerned, this subject is closed, at least until supporting evidence, of which there is currently /none/, is provided. The rest of you feel free to waste your time on this subject if you desire.

to pu benji ti fo lo mi me la.android. samcku toi

mu'o mi'e.aionys.

[John E Clifford]

Hard to argue with this except to raise an eyebrow at "well-known, established benefits (i.e., ease of learning).", where "widely claimed but often disputed and never tested" seems more accurate (a problem since at least 1960, when testing was proposed).

---

From: Jonathan Jones <eye...@gmail.com>
To: loj...@googlegroups.com
Sent: Thursday, June 21, 2012 9:53 AM
Subject: Re: [lojban] Re: Are Natlang the best case for entropy in communication ?

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

[tijlan]

On 21 June 2012 11:49, Escape Landsome <esca...@gmail.com> wrote:
> Suppose for instance we are given a new consonant "q", one could state
> that from now there is a strict equivalence between the phonological
> sequences
>
> "a" and "aqa"
>
> "e" and "eqe"
>
> "i" and "iqi"
>
> "o" and "oqo"
>
> "u" and "uqu"
>
> Thus, so'a and so'e can be confused in a noisy environement, but
> saying so'aqa ou so'eqe would avoid that.

A simpler solution:

: <- [ː]

a:
e:
i:
o:
u:
so'a:
so'e:
so'i:
so'o:
so'u:
...

Extend the vowel so that it can be heard out of the noise.
Additionally: speak louder.

If this doesn't work, then the problem isn't the language but either
the speaker's own vowel pronunciation or the surrounding noise itself.
Chances are, in such an overwhelmingly noisy environment, even your
"q" would be misheard as a different consonant: "aqa" as "aka", "iqi"
as "iji", etc. Consonants are no less mistakable than vowels. (I
regularly participate in my Japanese peers' Lojban workshop on Skype,
and not uncommon are confusions between /l/ & /r/, /b/ & /v/, /j/ &
/dj/, /z/ & /dz/, and /'/ & /x/, even when the phone line is fairly
clear. They could have complained about this and proposed a reform
analogous to yours, but they don't and instead just roll with it, keep
practicing.)

> We have no reason to get upset because on ONE PARTICULAR point
> natlangs behave better than lojban.
>
> There is even a chance that lojban can be amended in a way it behaves
> better then natlangs after the amendment.

> [...]

> This is a simple example to show you that discussing a drawback of
> lojban does not mean being mean towards it, rather it is what is
> expected from anybody here : that is, being scientific and examine
> closely and open-mindedly any problem.
>
> [I don't think the solution I proposed is a good one, either.   But at
> least it shows this is no dead-end street question.   And also we've
> no need to be aggressive.   Meanwhile, natlangs are still better than
> lojban on the entropy topic]

I don't think any irrational conservatism has been involved at least
in my responses to your idea. I'm actually rather leftist in this
community -- I can be in favor of any well-founded progressive
proposals aimed at improving Lojban or even creating a better
offshoot.

You are yet to substantiate your claim that {so'a/so'e/...} etc. are a
real problem. You have to explain how "a noisy environment" is or will
be so common a condition that a phonological/morphological reform is
necessary.

(Please note that I'm not angry at you. Don't be discouraged from
discussing your concerns about Lojban with other jbopre.)

mu'o

[Escape Landsome]

> If this doesn't work, then the problem isn't the language but either
> the speaker's own vowel pronunciation or the surrounding noise itself.
> Chances are, in such an overwhelmingly noisy environment, even your
> "q" would be misheard as a different consonant: "aqa" as "aka", "iqi"
> as "iji", etc.

I am misunderstood. "q" is an hypothetical new consonant. It has
nothing to do with guttural qâf, or any guttural, hence any "k" or
"kh" or "q". It is just a void symbol like "x" in algebra.

> A simpler solution:
>
> : <- [ː]
>
> a:
> e:
> i:
> o:
> u:
> so'a:
> so'e:
> so'i:
> so'o:
> so'u:
> ...

This does not work, nor does "VqV" either.

Why ? Because, suppose somebody mispronounces the "a" and "e" vowels.
This would indeed be a source of noise. What then ? She will say
"æ" instead of both "a" and "e".

But then, saying "æqæ" or "æ:" instead of "æ" does not help.

Thus, another kind of solution is needed !

[gleki]

That's exactly what I was saying. If you remove the difference between e/a you might go on with phoneme inventory reduction and end up in a language with only {a/i/u} and {p,t,k,m,n,s,h}

[Escape Landsome]

A noisy channel would not prevent somebody from understanding
correctly : "I want nune of those", or "I wand ell of those". But
the same noise applied to one of the so'V words would be fatal.

Anyway, this seems to be the price to pay for the easiness to learn to so'V set.

[tijlan]

On 22 June 2012 14:21, Escape Landsome <esca...@gmail.com> wrote:
>> If this doesn't work, then the problem isn't the language but either
>> the speaker's own vowel pronunciation or the surrounding noise itself.
>> Chances are, in such an overwhelmingly noisy environment, even your
>> "q" would be misheard as a different consonant: "aqa" as "aka", "iqi"
>> as "iji", etc.
>
> I am misunderstood.  "q" is an hypothetical new consonant.  It has
> nothing to do with guttural qâf, or any guttural, hence any "k" or
> "kh" or "q".   It is just a void symbol like "x" in algebra.

I'm not saying it has to do with those consonants. My point is: if the
listening environment was so bad that I would mishear some properly
pronounced vowel as a different vowel, I could also mishear some
properly pronounced consonant as a different consonant.

>> A simpler solution:
>>
>> : <- [ː]
>>
>> a:
>> e:
>> i:
>> o:
>> u:
>> so'a:
>> so'e:
>> so'i:
>> so'o:
>> so'u:
>> ...
>
> This does not work, nor does "VqV" either.
>
> Why ?  Because, suppose somebody mispronounces the "a" and "e" vowels.
>  This would indeed be a source of noise. What then ?   She will say
> "æ" instead of both "a" and "e".
> But then, saying "æqæ" or "æ:" instead of "æ" does not help.
>
> Thus, another kind of solution is needed !

The problem is her pronunciation. The first obvious solution is some exercise.

On 22 June 2012 15:42, Escape Landsome <esca...@gmail.com> wrote:
> A noisy channel would not prevent somebody from understanding
> correctly : "I want nune of those", or "I wand ell of those". But
> the same noise applied to one of the so'V words would be fatal.

- I want few of them.
- I wanted few of them.
- I want a few of them.
- I wanna feud over them.
- I won't feud over them.

More examples can be found in YouTube's automatic transcription:

http://www.youtube.com/watch?v=LifOwlzugbk

Turn on the CC button for the computer-generated subtitle, which is
supposed to represent the human speech as objectively heard. Note
where "when our economy was improving" is misheard as "when are
condoms improving" at 0:30, which I don't think is less fatal than
{so'o} being mistaken for {so'u}.


mu'o

[Jonathan Jones]

It's unfortunate that YouTube doesn't have a Lojban transcription. It would be interesting to see how well it transcribes us.

Moreover, it would would provide evidence for this very matter. Whether it would be for or against is still up for debate, however.

--
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.

[Escape Landsome]

>> A noisy channel would not prevent somebody from understanding
>> correctly : "I want nune of those", or "I wand ell of those".   But
>> the same noise applied to one of the so'V words would be fatal.
>
> - I want few of them.
> - I wanted few of them.
> - I want a few of them.
> - I wanna feud over them.
> - I won't feud over them.

Your objection is valid, yet it lacks some quantitative criterium.
Of course, it is true that given enough noise, anything can be
confused with anything else. The interest of studying entropy is
precisely to quantify the amount of noise under which error can be
avoided.

That is to say, this noise volume is much less in the case of a whole
confusible paradigme (as so'V is) than in any natlang case you would
give, or artifically create, for a paradigm is something that shares
both naturalness and commonness.

[Jonathan Jones]

On Sun, Jun 24, 2012 at 5:55 AM, Escape Landsome <esca...@gmail.com> wrote:

That is to say, this noise volume is much less in the case of a whole
confusible paradigme (as so'V is) than in any natlang case you would
give, or artifically create, for a paradigm is something that shares
both naturalness and commonness.

You keep saying that, yet still you provide nothing to support your claim. No evidence of any kind. It's beginning to sound like a mindless chant.

As those in the scientific community are quite fond of saying, PROVE IT.

[Escape Landsome]

> No evidence of any kind. It's beginning to sound like a mindless chant.
>
> As those in the scientific community are quite fond of saying, PROVE IT.

Entropy is defined as proportional to sum of p.ln p where p are the
probabities of the given choices. That means if you begin by
admitting equiprobability for the sake of simplicity, the more choices
you have the more confusion you get.

A paradigm being a natural set of choices that arises in natural
context, it follows that the complexity of two different paradigms
cooccurring in the same utterance is less than one single paradigm
which has product cardinal of the two former paradigms, because the
context specifies clues that helps to desambiguate each of the two
former paradigms by the other, --- which cannot be done with one
single big paradigm.

This is well-known and widely acknowledge in Linguistics.

Thus, the example you give with
[paradigm A with 2 choices] with [paradigm B with 2 choices]
is much less likely to confuse than the one with
[paradigm C with 2x2 = 4 choices]

the negative difference is the amount of information given by the cross-context.

Thus, any natlang example of 4 choices given as Aa, Ab, Ba, Bb, will
be less harmful than the lojban example of so'V where you have Xa, Xe,
Xi, Xo.

This general reasoning must be amended by the fact that some paradigm
are much more used than others. It is a reasonable argument that the
QUANTITY paradigm is much more useful than another very particular
paradigm, such as RACE OF HORSE, for instance.

[Jonathan Jones]

That is not evidence.

--
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.

[Escape Landsome]

Considering that nobody has metrics for Lojban or for the general case
of paradigms in totally any language, you won't have more than
qualitative argument based on mathematical properties of complexity
measures and generic minorations/majorations of allowed values.

Thus, the genetic argument is roughly the one given here --->
http://bluemoon.lescigales.org/Kholok/CollectiveDesambiguation.html

[Jonathan Jones]

No. No more argument. If you do not provide /evidence/, you will /never/ convince me.

I don't believe you. I have heard you repeat your argument time and again. In fact, all you've done /at all/ in this thread is repeat your argument, over and over again.

You have not once provided one reason for why you are even /talking/ about this, as Lindar asked you to do more than once, nor have you provided even the merest shred of evidence.

I don't need to hear your argument. I don't need to hear other arguments about similar things made by other people. This is not philosophy: we require proof.

--
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.

[Escape Landsome]

> No. No more argument. If you do not provide /evidence/, you will /never/
> convince me.

I just noticed that 10 minutes ago I've written this :

> Considering that nobody has metrics for Lojban or for the general case
> of paradigms in totally any language, you won't have more than
> qualitative argument based on mathematical properties of complexity
> measures and generic minorations/majorations of allowed values.

> Thus, the genetic argument is roughly the one given here --->
> http://bluemoon.lescigales.org/Kholok/CollectiveDesambiguation.html

Now, << the genetic argument >> was mistyped ! I meant << the geneRic
argument >>.

But everybody here understood me well. Why ? Because either the
difference between both words "genetic" and "generic" is slight in
context, or you had correctly CORRECTED me.

That is to say, the choice between X = { generic, genetic } has low
entropy in THIS utterance.

We could study other examples where a single mistyping of a single
letter would happen.

For instance in French : << et alors elle a ouvert la mouche pour
parler >> is correctly understood as << la bouche >> instead of << la
mouche >>. Still the set X of all possible choices is big : X = {
bouche, couche, douche, louche, mouche, souche, touche }

But the context prevents other Xouche words than body organs to be
selected, so the entropy is very low in context.

-

Now consider a lojban sentence where somebody would talk of a number
of people in a room, he would use so'V but would mistype the V letter.

The X set has 5 elements but none of them can be best selected given
the context.

Hence we conclude the entropy would be high in context.

-

If we consider all these examples, we see the last example entropy is
high because of the gathering of VERY OFTEN-USED FEATURES in a VERY
NARROW SET of VERY SIMILAR FORMS, which is not the case of the former
examples, since for those the paradigms are not natural.

Hence you see the lojban case has very high entropy while the two
natlang examples have not. If this is not so, GIVE ME A NATLANG
EXAMPLE WITH NATURAL PARADIGM AND 5 ALTERNATIVES OR MORE, given the
same condition of mistyping of A SINGLE LETTER.

The difference of greatness in entropy in lojban and natlang examples
is so flagrant you don't even need to measure it, as if we were
speaking of the size of an elephant compared to the size of a mouse.

[Jonathan Jones]

Still argument without evidence. You are theorizing, and drawing conclusions from this UNTESTED theory. EVIDENCE OR GTFO.

--
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.

[Escape Landsome]

* NUMERICAL ESTIMATE OF ENTROPY IN EXAMPLE (1) :

For instance in French : << et alors elle a ouvert la mouche pour
parler >> is correctly understood as << la bouche >> instead of << la
mouche >>.   Still the set X of all possible choices is big : X = {
bouche, couche, douche, louche, mouche, souche, touche }

But the context prevents other Xouche words than body organs to be
selected, so the entropy is very low in context.

p(bouche) > 0.9, sum p(Xouche) < 0.1 for other X, since worst case is
equiprobability for all those,

S(example 1) < -(0.9*log_2(0.9) + 6*0.1*log_2(0.1)) = 2.129


* NUMERICAL ESTIMATE OF ENTROPY IN EXAMPLE (2) :

Now consider a lojban sentence where somebody would talk of a number
of people in a room, he would use so'V but would mistype the V letter.

The X set has 5 elements but none of them can be best selected given
the context.

THEN S(example 2) ~ -5*log_2(0.2) = 11.609


* 11.609 IS MUCH BIGGER THAN 2.129

[Jonathan Jones]

That's not evidence.

/ignore

--
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.

[Escape Landsome]

> That's not evidence.
>
> /ignore

You /ignore so well that you didn't notice I made a miscalculation.

The correct value are :

S(example 1) neatly inferior to 2.129 by brutal majoration, equal to
0.709 with highest probability

S(example 2) equal or very near to 2.321

A value of 11 was incredibly high. But 2.321 is still vey superior to 0.709

[Jonathan Jones]

EVIDENCE is date gathered from experiments designed to support or refute a hypothesis.

Where is the record of the experiment you performed in order to acquire this data? Nowhere, because you didn't do any experiments and are therefore /making it all up/.

Shit-you-made-up is never evidence.

--
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.

[Jonathan Jones]

On Sun, Jun 24, 2012 at 8:35 AM, Jonathan Jones <eye...@gmail.com> wrote:

EVIDENCE is date gathered from experiments designed to support or refute a hypothesis.

*data

[Escape Landsome]

> EVIDENCE is date gathered from experiments designed to support or refute a
> hypothesis.

* In hard science like Mathematics or Physics, yes, but those do not
have the monopoly of scientificity, and Linguistics being a much more
"human social science", accepts scientific argument in a more laxist
way.

In social sciences, probability is accepted under the subjective form
it has, concurrently to objective measurement it has too (if you go
read the articles on Probability theory in Wikipedia or Britannica,
you'll see the foundations of probability theory accepts both
interpretation of probability, and it is still discussed which one is
the true one)

* So, first, in the case of Example 2, since the utterance says how
many people there are, and it is impossible to tell outside of the
utterance, there is equiprobability, and this is proved. Thus
S(Example 2) = -5*(0.2)*log_2(0.2) = 2.32 is *PROVED*

If you are not convinced, or if you look for a qualitative reasoning
that confirms it, think that there are thus 5 equiprobable
possibilities, so you need to code {a,e,i,o,u} in binary digits, and
you need a little more than 2 bits for that, 2.32 is then the good
result.

* For the Example 1, "bouche" is highly probable, and the 6 others
highly strange. Let P be the probability of the "probable event",
that is "X = b in bouche".

Then S(Example 1) is -P.log_2(P) + other terms

where other terms is between P.log_2(1-P) and (P/6).log_2((1-P)/6) [as
can be proved mathematically]

The value of P should be computed from a corpus, but without that we
can bound it by minoration and it is certainly > 0.8, so, do the
calculations, and you'll see this holds.

* Anyway, you want something else, that nobody can give you at the
present, that is (unnecessary) precise studies, and even then, I bet
you won't be convinced. What would convince a believer ?

[Jonathan Jones]

On Sun, Jun 24, 2012 at 8:49 AM, Escape Landsome <esca...@gmail.com> wrote:

> EVIDENCE is date gathered from experiments designed to support or refute a
> hypothesis.

* In hard science like Mathematics or Physics, yes

No, in ALL science. If there is no experimentation, it ISN'T science.

While it is true that the so-called "soft" sciences such a psychology, sociology, et al., are usually incapable of coming to concrete conclusions, they still use experiments to accrue evidence of their hypotheses.

Your mathematical formulae have no basis in any kind of experimenting, and thusly can not be given any higher regard than numbers which are completely made up. You must have experiments to back up your claim, for without them, that's all you have; a claim.

Furthermore, even after repeatedly being requested to do so, you have yet to provide a reason for why you've brought this up in the first place.

I am not a "believer". I hold my opinion for the simple reason that there is no evidence to the contrary.

[Jonathan Jones]

Also "scientificity" isn't a word.

[Jonathan Jones]

On Sun, Jun 24, 2012 at 8:49 AM, Escape Landsome <esca...@gmail.com> wrote:

* Anyway, you want ... (unnecessary) precise studies....

No, what I want is studies, period. At this point, there have been no case studies regarding the relative likelihood of miscomprehension between utterances in Lojban and Natlang-X, which means there is absolutely no evidence one way or another about the subject. As I said before, this isn't philosophy. Without evidence, there can be no conclusions, only claims and arguments.

[Escape Landsome]

> Also "scientificity" isn't a word.

It is in my natlang, but sorry this does not exist in English.



Come, what you want is the prove that S(Example 2) much bigger than
mean value of S(Example 1s) for all Example 1s we can imagine.

I reply with a minoration. A minoration IS A MATHEMATICAL proof.
If I was asked by you if an elephant is bigger than a mouse, even
without precisely measuring both animals, I could reason like that :

Size(Elephant) much bigger than 1 meter

Size(Mouse), smaller then 0.30 meter

=> thus Size(Elephant) much bigger than Size(Mouse)


This is the same here, simply you don't perceive how much bigger is
S(E2) relatively to all S(E1).


But, anyway, well, I surrender. There must be somewhere studies
recording estimates of S(E1s) for a lot of E1 situations where one
letter is mistyped.

And I bet mean value of S(E1s) is between 1 and 1.58, which is the
most natural estimate tells us that in natural context there are in
the worst cases 2 ou 3 possibilities.

You see, I make this a prediction that can be falsified, in a
popperish way of science. This is what I bet.

And this is still less than 2.3 which is THE PROVED VALUE of S(E2).

[Jonathan Jones]

You can't do a mathematical proof when dealing with humans. Humans are much too complex to be able to be reduced to mathematical forms, especially with regard to social interactions, of which language is a member.

And you STILL HAVEN'T SAID WHY YOU BROUGHT THIS UP.

--
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.

[Escape Landsome]

What is wanted from an engineer is not only to measure things or make
correct proofs of theorem but also to have a right sense of magnitudes
of values.

That is, when I studied physics (and not philosophy, as you would have
guessed), my teachers asked me to prove that no human would experiment
gravity in a hollow Earth, or that some satellite would fall with the
speed V, but *also* they saked tricky questions, such as : "what is
your volume ?", and you had to answer "roughly 70 liters", since you
reason qualitatively thus :

"my body is mainly water (more than 80%) and 1liter of water is 1kg,
and I weigh 70kg, so my countenance is 70 liter"

This is qualitative intuitive estimation, and is accepted as good in
science and engineering !

-

I deeply regret that you have no INTUITION of what linguistic entropy
is, for if that were the case, you would see as SOMETHING OBVIOUS that
S(lojbanic example) far exceeds S(example in natlang)

One acknowledged interpretation of linguistic entropy is that when you
have approximately n responses that could do the guess, and all k
others that are too rare or too strange, as the computation shows, the
k contributions are negligible, and roughly the mean value will be the
same as a choice between n choices.

Since then, you see by immediate intuitive estimation, analogous to
the ones you would do in engineering, that mean value of S must lie
between log_2(1) an log_2(3), for the most often cases have between 1
and 3 possible alternatives, and more than 3 is rare in a paradigm.

Thus, mean value is between 0 and 1.58.

This is rough estimation but this is science.

-

But you insist on having (unnecessary) precise study, this is rather
boring, since you are in the same posture that somebody would would
not accept that the distance from Paris to the Moon is far greater
than from Paris to Moscow, since he hasn't a PRECISE value.

You insist on having precise values instead of estimation of
magnitudes, when the latter suffices vastly. This is a way to make
yourself boring, and to lock to debate with the kind of trick somebody
could use in a debate where he doesn't accept the evident truth.

[Jonathan Jones]

On Sun, Jun 24, 2012 at 9:38 AM, Escape Landsome <esca...@gmail.com> wrote:

But you insist on having (unnecessary) precise study....

No. Again, as I said, I insist on HAVING A STUDY AT ALL. If you do not do experiments, then all you have is a hypothesis. Your formulae are no more than predictions based upon that hypothesis. You have NO PROOF, and NO EVIDENCE. You ONLY have a CLAIM.

I refuse to participate in this circular and increasingly obviously pointless debate with you any further.

You don't tell us WHY you've brought this subject up, you REFUSE to do anything more than CONTINUOUSLY and REPEATEDLY make the SAME ARGUMENT with NO EVIDENCE to SUPPORT YOUR CLAIM.

You, sir, are no scientist, and I will not be dragged any further into this foolhardy endeavor you waste our bandwidth on. Good day.

I SAID GOOD DAY!

[.arpis.]

aionys is pointing out that even though, mathematically, you can find
a difference in ambiguity, this doesn't (necessarily) mean anything in
practice. There are a whole host of things that can go wrong between
the model and the real world (the model may imperfectly model the real
world, the model may perfectly model something that is not the real
world, etc.)

Also, taking for a given that lojban has more potential ambiguity due
to ambient noise or mishearing, that doesn't mean that _communication_
in lojban is any more difficult than in other languages; discourse
might structure itself so that this is not an issue.

Even taking for granted that communication in lojban is more difficult
than in other languages, you have failed to answer the question that
other people have posed: "so what?"

Do you propose a change to lojban? It's not going to happen, as it
will have to be so pervasive as to invalidate all the lojban we
already know.
Do you propose a new language? Design it yourself and come back to us;
we may not learn it, but we'll appreciate your work (and point out
flaws in it that you hadn't thought of).
Do you propose we just give up on lojban? You're posting this to the
lojban mailing list; you can imagine the outcome yourself.

One of the strengths of lojban, entirely apart from its design, is
common to most successful open source projects; we are willing and
able to avoid yak shaving and bike shedding, and even though we
sometimes enjoy a jbodau here and there, we can completely ignore
proposals without feeling bad. there are simply too many changes to
consider, and stability is more important.

> --
> 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.
>

--
mu'o mi'e .arpis.

[la .lindar.]

Escape Landsome, why are you bringing this up?

[Michael Turniansky]

Arpis, thank you for your reasoned and moderate tone.  Unlike Lindar and aionys, whose version of dialogue is simply sticking their fingers in their ears and saying "lalala, I'm not listening, you're an a**hole" (really?  THIS is the attitude we want to have towards people who have an interest in our language?), you are actually explaining their concerns to Escape.

Aionys, you keep harping about experimentally proving this, rather than the self-evident mathematical proofs.  So tell me -- how would YOU design an experiment to prove or disprove the hypothesis? You design it, we can carry it out.

li'a lo ni frica zo so'a zo so'u na barda .ije da'i do na krici du'u go'i .ibo mi stidi lo nu do sanli ca'u le julne tanxe fau lo nu da lausku lu ko tolri'ugau so'* lo vu cilce jo'u xagji cinfo li'u do

(and personally, I've always wondered why the CLL makes such a big deal about the digits being easily told apart in noisy environments (18.2) when as clearly demonstrated here and in so many other places (ko'V series, fV series, etc.), it's not the case.  Better for the CLL to not make the claim at all, since it just sets up its own counterarguments in other area of the language (my personal opinion when I first read that passage 8 years ago?  It was simply a dig at JCB and Loglan, which uses a different system which is much easier to memorize for the beginner (cf. tiljan and gleki's arguments about the matter at hand) )

Escape's motivations for asking a question should not be an issue ("Appeal to Motive" fallacy).  And attacking him for the question simply makes him  defensive, as it would any person.  Why can't a person simply point out a language feature that bothers them without being pilloried?  A simple, "Yes, we know about it, but we're not likely to change it, and we don't think it to be an actual problem in real life situations more than any other natlang" would have made this thread a lot shorter.

Now, as to Escape's contention that such kind of phonemic ambiguity in words of potentially disastrous confusion (i.e., opposite words in the same scale) doesn't exist in natural languages, I WILL challenge that assertion.  <facetiousness> Really?  Have you studied all the scales in all of the world's many thousands of languages?  You impress me!  </facetiousness> I haven't found any yet, but I can't dismiss the possibility it exists in a natlang...

\                   --gejyspa

[Jorge Llambías]

On Fri, Aug 10, 2012 at 10:43 AM, Michael Turniansky
<mturn...@gmail.com> wrote:
>
> Now, as to Escape's contention that such kind of phonemic ambiguity in words
> of potentially disastrous confusion (i.e., opposite words in the same scale)
> doesn't exist in natural languages, I WILL challenge that assertion.
> <facetiousness> Really? Have you studied all the scales in all of the
> world's many thousands of languages? You impress me! </facetiousness> I
> haven't found any yet, but I can't dismiss the possibility it exists in a
> natlang...

How about "can" vs "can't" in English? The difference between them can
be rather minimal.

mu'o mi'e xorxes

[And Rosta]

Michael Turniansky, On 10/08/2012 14:43:

> (and personally, I've always wondered why the CLL makes such a big
> deal about the digits being easily told apart in noisy environments
> (18.2) when as clearly demonstrated here and in so many other places
> (ko'V series, fV series, etc.), it's not the case. Better for the CLL
> to not make the claim at all, since it just sets up its own
> counterarguments in other area of the language (my personal opinion
> when I first read that passage 8 years ago? It was simply a dig at
> JCB and Loglan, which uses a different system which is much easier to
> memorize for the beginner (cf. tiljan and gleki's arguments about the
> matter at hand) )

I don't understand your point. With the exception of re/rei, the digits *are* maximally distinct, and that is a virtue, especially in lexical domains where context is unlikely to be able to disambiguate, such digits and letters. We see in English that _x-ty_ and _x-teen_ words are frequently replaced by _x-zero_ and _one-x_, and that on the telephone the alpha-bravo-charlie-delta system is used for letter names. It's true that other series aren't internally maximally distinct, but partial internal sameness enhances learnability, and the ko'V and fV series at least make use of vocalic contrasts, which are acoustically more salient than consonantal ones (tho for reasons of acoustic distinctness, ko'V would better have been kV'o).

> Now, as to Escape's contention that such kind of phonemic ambiguity
> in words of potentially disastrous confusion (i.e., opposite words in
> the same scale) doesn't exist in natural languages, I WILL challenge
> that assertion. <facetiousness> Really? Have you studied all the
> scales in all of the world's many thousands of languages? You impress
> me! </facetiousness> I haven't found any yet, but I can't dismiss the
> possibility it exists in a natlang...

In many accents of American English, _can_ and _can't_ are very similar or even sometimes identical, and it is a known impediment to effective communication.

A slightly different phenomenon is autoantonyms (<http://en.wiktionary.org/wiki/Appendix:Glossary_of_auto-antonyms>), most of which involve polysemous words with opposite polysemes. Some of these (e.g. apology, sanction) are also not always disambiguable by context and hence are impediments to effective communication.

--And.

[MorphemeAddict]

On Tue, Jun 19, 2012 at 4:53 AM, Escape Landsome <esca...@gmail.com> wrote:

> I can think of no case where "very different concepts are given very near
> phonological forms". It is true that there are many cmavo which are very
> nearly the same, such as FA, SE, etc., but all of these groupings are very
> closely related to each other, differing in very minor ways.
>
> Your own example of the so'V series is not nearly as "different" as you seem
> to think. Each of them is a point on the scale All-None, exclusive. The only
> difference is where on that scale each is, with so'a being closest to All,
> and so'u closest to None.

The paradigm is the same (all so'V are related to scale All-None) but
inside this paradigmatic choice, they are very distinct.

All is not at all the same thing as None or as Few, and if somebody is
not well heard when saying the word, the consequences are great !

As if I said [ x = 0% ], [ x = 10% ], [ x = 100% ], and all the onus
of communication lied on the real variable near the percent sign...
Communication would be in great risk to be lost.

In the case of natlang some redundancy is set to avoid this, namely
the "f-" or "few" opposes the "n-" of "none", but also "-ew" of "few"
opposes "-one" of "none", so that, if ever one phonem is not well
understood, the other ones are there to save the day.

>> And : << Does not the fact natlangs do not have this problem generally
>> speaking imply that they are more well designed than lojban on this
>> particular point ? >>
>
>
> I don't think so, mainly because, being NATlangs, they weren't designed at
> all. I highly doubt that the phonological properties of words were ever
> taken into account during the etymological evolution of those languages. In
> fact, there is ample evidence to the contrary, as there are many cases where
> words that are /extremely/ different concepts have /extremely/ similar
> phonotactics, such as the English to, too, and two. I would argue that
> mistaking, for example, "bow" (either the act-of or the ship-part, but not
> the clothing) and "now", or "to" and "too" is much more damaging to listener
> comprehension than mistaking "so'a" for "so'u".

Natlangs were designed, but the designer is a non-human (and
non-divine) random process of natural selection.

A random process of natural selection is a good definition of NON-design. 

stevo

  Natural selection
favorises random creations, but random creations naturally occupy the
phonological space smoothly and in a sparse way, so one can say,
natural selection naturally designs words that are good for efficient
communication.

[MorphemeAddict]

On Mon, Jun 18, 2012 at 5:52 AM, Escape Landsome <esca...@gmail.com> wrote:

Coi rodo,

in a noisy verbal interaction, for instance, on a phone call (but this
applies to any case, indeed), it is good for communication that the
elements of signal can be easily distinguished, so as to avoid
reception errors.

This is achieved by natlangs, by sparsing all the existing words (for
instance, adverbs) in such a "morphologic space" that has very few
collisions.   For instance, the english words for "few", "many", "a
lot" and "none" are phonologically very different of each other, so
there is little chance you could confuse them by hearing them on a
deficient phone.

But this is not the case of lojban words, for instance so'a, so'e,
so'i, so'o etc. are very near of each other, and, assuming you don't
hear well the last vowel, you could infer something very far from what
was intended by the other speaker.

So, is not that something that is annoying ?

Like French "dessous" ('underneath') vs. "dessus" ('over'). The only difference is [u] vs. [y]. Apparently the French have no problem with it (or do they?), but it seems like a big problem to me. 

stevo 

[Pierre Abbat]

On Tuesday 14 August 2012 12:30:49 MorphemeAddict wrote:
> Like French "dessous" ('underneath') vs. "dessus" ('over'). The only
> difference is [u] vs. [y]. Apparently the French have no problem with it
> (or do they?), but it seems like a big problem to me.

Doesn't seem like a big problem to me, though I once made a tonguetwister out
of similar words: "La roue sur la rue roule; la rue sous la rue reste." I
know several semantically similar minimal pairs in natlangs:
hu: tizenegyedik (eleventh)/tizennegyedik (fourteenth). This requires both
using the ordinal (four is négy, with a long vowel) and adding a multiple of
ten to get the double n (and first is első).
es: sesenta (sixty)/setenta (seventy). This does cause me problems.
fr: il (he)/elle (she)
pt: ele (he)/ela (she)
en: he/she. Unlike the preceding two pairs, these are unrelated.
Tz'utujil: ka'- (two)/kaj- (four) and waq- (six)/wuq- (seven). The root for
eight is waqxaq, which may lead to further confusion.
I don't know of semantically similar quints in natlangs.

Pierre
--
I believe in Yellow when I'm in Sweden and in Black when I'm in Wales.

[Michael Turniansky]

On Tue, Aug 14, 2012 at 11:14 AM, And Rosta <and....@gmail.com> wrote:

Michael Turniansky, On 10/08/2012 14:43:

(and personally, I've always wondered why the CLL makes such a big
deal about the digits being easily told apart in noisy environments
(18.2) when as clearly demonstrated here and in so many other places
(ko'V series, fV series, etc.), it's not the case. Better for the CLL
to not make the claim at all, since it just sets up its own
counterarguments in other area of the language (my personal opinion
when I first read that passage 8 years ago? It was simply a dig at
JCB and Loglan, which uses a different system which is much easier to
memorize for the beginner (cf. tiljan and gleki's arguments about the
matter at hand) )

I don't understand your point. With the exception of re/rei, the digits *are* maximally distinct, and that is a virtue, especially in lexical domains where context is unlikely to be able to disambiguate, such digits and letters. We see in English that _x-ty_ and _x-teen_ words are frequently replaced by _x-zero_ and _one-x_, and that on the telephone the alpha-bravo-charlie-delta system is used for letter names. It's true that other series aren't internally maximally distinct, but partial internal sameness enhances learnability, and the ko'V and fV series at least make use of vocalic contrasts, which are acoustically more salient than consonantal ones (tho for reasons of acoustic distinctness, ko'V would better have been kV'o).

   I am sorry you don't see my point.  I will try to clarify.  I never said that the digits aren't maximally distinct (although in addition to rei/re confusion, "pi" can easily be confused with bi (and to a lesser extent, pa) in a noisy environment.)  But you make two claims above -- A) "[maximal distinctiveness]  is a virtue, especially in lexical domains where context is unlikely to be able to disambiguate, such digits and letters" and B) "partial internal sameness enhances learnability"  These are contradictory claims. Which is the virtuous one?  In original Loglan, the digits 0-9 were ni, ne, to, te, fo fe, so, se, vo, ve, which directly addresses point B,but in lojban we chose to chose A over B WITH REGARD TO DIGITS, and tout it in the CLL with a direct call-out.  In letters (and in many other selma'o) , on the other hand, we chose option B, despite the fact that "letters" is one of the places that you say A is better suited.  And in fact, cmavo space is so tight, confusion will always be a problem in noisy environments ("Did he say "lo pa jatna cu morsi" or "lo ba jatna cu morsi"?")  The big answer is "So what?  Nothing we can do about it to satisfy all situations.  Humans have evolved to deal with ambiguity in communication, inlcuding "ki'a/ke'o.  So why draw attention to it in the CLL?" was my only point here.

     --gejyspa

[And Rosta]

Jorge Llamb�as, On 11/08/2012 01:36:

Sorry to xorxes; as I wade through vast email backlog I find xorxes had already made this observation before me.

--And.

[And Rosta]

Michael Turniansky, On 14/08/2012 18:16:

> I am sorry you don't see my point. I will try to clarify. I never
> said that the digits aren't maximally distinct (although in addition
> to rei/re confusion, "pi" can easily be confused with bi (and to a
> lesser extent, pa) in a noisy environment.) But you make two claims
> above -- A) "[maximal distinctiveness] is a virtue, especially in
> lexical domains where context is unlikely to be able to disambiguate,
> such digits and letters" and B) "partial internal sameness enhances
> learnability" These are contradictory claims. Which is the virtuous
> one?

They're both virtues, but, as is often the case with virtues, not fully compatible. For the digits, I think Lojban is right to prioritize distinctiveness over learnability, tho something like {ki ke ka ko ku mi me ma mo mu} would also have worked equally well (i.e. not varying the consonant for every digit, and choosing very distinct consonants). Elsewhere, I think (as I said in my last msg) the pattern of keeping the consonant the same within paradigms and varying the vowel is a good one.

>In original Loglan, the digits 0-9 were ni, ne, to, te, fo fe,
> so, se, vo, ve, which directly addresses point B,but in lojban we
> chose to chose A over B WITH REGARD TO DIGITS, and tout it in the CLL
> with a direct call-out. In letters (and in many other selma'o) , on
> the other hand, we chose option B, despite the fact that "letters" is
> one of the places that you say A is better suited.

Sure: just as credit is due for the digit names, debit is duefor the letter names. I don't think that means CLL should keep shtoom about the credit for the digit names, tho.

> And in fact, cmavo
> space is so tight, confusion will always be a problem in noisy
> environments ("Did he say "lo pa jatna cu morsi" or "lo ba jatna cu
> morsi"?") The big answer is "So what? Nothing we can do about it to
> satisfy all situations.

Of course there's something that can be done about it! Ensure that words that context will be least likely to be able to disambiguate have the more distinct forms. That minimizes the chances of mishearing.

> Humans have evolved to deal with ambiguity in communication, inlcuding "ki'a/ke'o.

As I said in my last, when context doesn't disambiguate, the language changes.

>So why draw attention to it in the CLL?" was my only point here.

Because it's a good design feature. It's an area of the language that doesn't need to be fixed.

--And.