racli/ralci

[Jonathan Jones]

I do not like these two words. They are both two (defined) place gismu, they have very different meanings, and they differ ONLY in the order of c and l, which is problematic for people with dyslexia. I feel they should at least end with a different vowel. I would be very annoyed if I was to accidentally say for example that because glass breaks when hit with a hammer, it is mentally sound.

Just felt like sharing.

--
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 )

[la .lindar.]

kalci - kelci

ractu - ratcu

cirla - cilre

simsa - simlu

>___>

Just sayin'...

[Jonathan Jones]

None of those are as bad, from a dyslexic's standpoint.

Granted, plaything/crap is pretty awful to get mixed up, but it's bad because they only differ in one vowel. You can't get one from the other by rearranging the letters, which is what dyslexia is all about.

Rabbit/Rat is nearly as bad, but the meanings are very similar (a rat is a rodent, whereas a rabbit is rodent-like (they're first cousins or possibly even siblings genetic-tree-wise).

Learn/Cheese at least has a different ending vowel, so only really causes problems in the 4L-rafsi form, which isn't much of a porblem.

Seems/Similar are practically synonyms, and like kalci/kelci, their problem doesn't stem from potential dyslexic misspelling.

[Pierre Abbat]

On Friday 27 July 2012 22:19:42 la .lindar. wrote:
> kalci - kelci
> ractu - ratcu
> cirla - cilre
> simsa - simlu

garna - grana
junla - julne
simlu - simxu
cmaci - jmaji

The ones that keep confusing me are "ratcu/ractu" (both mammals, and
lagomorphs were classified within the rodents) and "grana/garna" (significant
overlap of extension). The others don't bother me as much. If someone said "I
put my goat on because it's gold", you can figure out what he meant.

Pierre

--
La sal en el mar es más que en la sangre.
Le sel dans la mer est plus que dans le sang.

[Jacob Errington]

No one raised fetsi/festi ? Or/tsani stani ? Or tinsa/tisna ? 

If we're going to play with voiced/unvoiced, why not gugde/kukte ? Or hey stani/zdani ?

Lojban is full of this and anyone with any kind of speech disorder/impediment/whatever-it's-called will probably have a bit more trouble than in a natural language, but of course Lojban has other benefits or advantages versus other languages, so it's all a matter of perspective, not to mention that it's hardly likely that this will ever be mixed. (The morphology seems rather set in stone.)

mu'o mi'e la stani

--
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.

[John E Clifford]

hile the morphology is set in stone, LoCCan3 (which, admittedly, does meditate on fiddling with the morphology, too) considered redoing the vocabulary, which is largely happenstantial, though with some thought given to various sorts of easy confusions.  Someone worked once -- I have no idea who or when, now, or where it might be stored -- the number of  CCV/CVCCV forms with all the Lojban constraints which would avoid all the obvious problems (though what is obvious varies from person to person, culture to culture,etc.) and came up with a number rather lower than the number of gismu now in use.  I remember that care was taken about voiced/voiceless, about simple order shifts in clusters, about /r/ vs both /l/ and /x/, but I think there were other factors.  Anyone remember -- or can suggest -- others?   Maybe CC/C exchanges?  And something about vowels, too, but I don't remember what.

---

From: Jacob Errington <nict...@gmail.com>
To: loj...@googlegroups.com
Sent: Sunday, July 29, 2012 8:33 PM
Subject: Re: [lojban] Re: racli/ralci

[Ian Johnson]

This is basically not fixable without moving the problem somewhere else. It's messy because there are so many distinct collision rules, but try going through and computing how many gismu can technically exist at the same time. It's not as big a number as you'd think. Without getting into the various rules introduced later in the CLL (which were meant to reduce problems like these), the number is still only around 22,000. Here's a way to get started:
5 vowels
17 consonants
48 initial clusters
{11 unvoiced consonants, 10 voiced consonants, (m, l, n, and r get double counted) 9 "special exceptions": cs sc jz zj cx kx xc xk mz are all forbidden, and no double consonants}->191 internal clusters
gives exactly 20315 gimtai. This avoids only collisions of final vowels, and permits every other apparent "collision", including all those listed in the CLL. These restrictions, to within a very crude approximation, probably divide the number given above by something between 2 and 3, giving a number between 6700 and 10000, probably closer to the former given that most of the collisions involve two letters. This is honestly pretty tight, to fit in ~1300 words.

I agree with you, though, that some fiddling with the gimste, doing things like changing final vowels to move further away from the source languages to avoid some collisions would probably have been for the best. Too late now, of course.

An interesting computation: assuming every combination of 2 vowels and 3 consonants can serve as exactly one gismu (which lets words differ only in their final vowel; not doing so makes the current number of gismu just barely fit), you get 6800 possible words. Almost all such combinations work, too: ignoring the 9 special exceptions, any 2 vowels obviously work, and any group of 3 consonants will have at least 2 with have the same voicing. It might be interesting to make up a hypothetical gimste subject to this restriction and none of the restrictions I linked (but still the voicing rules and the special exceptions).

mu'o mi'e la latro'a no'u la latros

[Ian Johnson]

Eh, s/22,000/20,000; I had forgotten about the double consonant rule, then changed it in one place and not the other.

mu'o mi'e latros

[MorphemeAddict]

The actual shape of the words (phonology?) is pretty much irrelevant. It's the grammar (syntax? morphology?) that matters. Once that's in place, the actual words can be added on. 

stevo

[MorphemeAddict]

Are we *ever* going to remove "mz" from the list of exceptions? It should never have been there in the first place. 

I'd vote for allowing "cx" and "xc" too. 

And while xk/kx are disallowed, pf/fp and ts/st are allowed. Consistency is important. No exceptions! 

stevo

On Wed, Aug 1, 2012 at 8:50 PM, Ian Johnson <blindb...@gmail.com> wrote:

[Jorge Llambías]

On Wed, Aug 1, 2012 at 9:50 PM, Ian Johnson <blindb...@gmail.com> wrote:
> 5 vowels
> 17 consonants
> 48 initial clusters
> {11 unvoiced consonants, 10 voiced consonants, (m, l, n, and r get double
> counted) 9 "special exceptions": cs sc jz zj cx kx xc xk mz are all
> forbidden, and no double consonants}->191 internal clusters
> gives exactly 20315 gimtai. This avoids only collisions of final vowels, and
> permits every other apparent "collision", including all those listed in the
> CLL. These restrictions, to within a very crude approximation, probably
> divide the number given above by something between 2 and 3, giving a number
> between 6700 and 10000, probably closer to the former given that most of the
> collisions involve two letters.

I don't think so. "blanu" for example will block "planu", but it won't
block "plana", so by playing with the final vowel you can probably
accomodate all 20315 four-letter initial forms.

> An interesting computation: assuming every combination of 2 vowels and 3
> consonants can serve as exactly one gismu (which lets words differ only in
> their final vowel; not doing so makes the current number of gismu just
> barely fit), you get 6800 possible words.

Three _different_ consonants and two _different_ vowels, right? But
you can have gismu with two repeated consonants or vowels (e.g.
"nanba"). So I think the number for different combinations comes out
as 14280.

> Almost all such combinations work,
> too: ignoring the 9 special exceptions, any 2 vowels obviously work, and any
> group of 3 consonants will have at least 2 with have the same voicing.

I would guess around 5-10% are blocked by the special exceptions.

mu'o mi'e xorxes

[Jonathan Jones]

I wasn't really proposing any changes. I was really just complaining. There's no way that the designers could have predicted every possible problem that could occur, and so obviously, as time goes on, we will find more and more. I'd be willing to bet that these problems will most likely evolve out eventually, either through various proposals as with xorlo, or through natural evolution when the language finally reaches the point of being a first-order language. (And by first-order, I mean the first language the peoples of least one culture learn, nothing more.)

[Ian Johnson]

On Wed, Aug 1, 2012 at 9:38 PM, Jorge Llambías <jjlla...@gmail.com> wrote:

I don't think so. "blanu" for example will block "planu", but it won't
block "plana", so by playing with the final vowel you can probably
accomodate all 20315 four-letter initial forms.

I see what you mean. It would be interesting to attempt this computationally; the "transitivity" if you will of the blocking rules makes this seem like it could be subtler than either of us are seeing. You're closer to right than I was, though, if not actually right.

> An interesting computation: assuming every combination of 2 vowels and 3
> consonants can serve as exactly one gismu (which lets words differ only in
> their final vowel; not doing so makes the current number of gismu just
> barely fit), you get 6800 possible words.

Three _different_ consonants and two _different_ vowels, right? But
you can have gismu with two repeated consonants or vowels (e.g.
"nanba"). So I think the number for different combinations comes out
as 14280.

 

Good point, I didn't think of that possibility. I actually just wrote some code that puts down a gismu list with all letters distinct. It's pretty readily extensible; in particular I can easily change the order of the nesting, which significantly alters the words that actually get generated. A variant in which initial consonants vary first, then the cluster, then the initial vowel, then the final vowel, with no initial clusters whatsoever, results in http://pastebin.com/SYfHQf8p, which I found pretty interesting.

> Almost all such combinations work,
> too: ignoring the 9 special exceptions, any 2 vowels obviously work, and any
> group of 3 consonants will have at least 2 with have the same voicing.

I would guess around 5-10% are blocked by the special exceptions.

Well, in the slightly naive computation earlier, it's 9/(191+48) =  about 3.8%.

[Ian Johnson]

Erm, woops, no, it doesn't force the letters to be distinct, as you can see from the very beginning :)

[John E Clifford]

Gleanings from yellowing LoCCan3 files.

# acceptable consonant frames (CC/C, C/CC) = 2975

# if only one of C1C2, C2C1 is allowed = 1836

# if only one of C1C2, C1'C2' is allowed (' being opposite voicing) = 1819

# if only one of C1C2/C3, C3/C1C2 is allowed = 2159

There are also notes about various combinations of these (cluster-singleton exchange with voicing and order reversal, say) but not numbers.

There are numbers for cases where only r or l can occur in a given place in a frame and similarly for the sibilants, but since these do nothing to solve the same problem in the cmavo, they are not mentioned here.

Notice, all these assume the standard initial cluster restrictions, which seem unnecessarily strict (would tf or dv be a problem, or tl or dl?  And there are probably others.  Oh, yes, medial mz).  Even so, all these further restrictions leave enough frames to allow every gismu to have its own frame and leave vowels to serve other redundancy increasing tasks (taking care of sibilants and r/l, for example), even if we have only a 3-way stressed vowel distinction and no distinguishing unstressed vowels.

Just fodder for the flames of the next generation (when that charismatic nutcase comes along).

---

From: Ian Johnson <blindb...@gmail.com>
To: loj...@googlegroups.com
Sent: Thursday, August 2, 2012 4:27 AM
Subject: Re: [lojban] racli/ralci

[Jorge Llambías]

On Thu, Aug 2, 2012 at 6:10 AM, Ian Johnson <blindb...@gmail.com> wrote:
> On Wed, Aug 1, 2012 at 9:38 PM, Jorge Llambías <jjlla...@gmail.com> wrote:
>>
>> I don't think so. "blanu" for example will block "planu", but it won't
>> block "plana", so by playing with the final vowel you can probably
>> accomodate all 20315 four-letter initial forms.
>
> I see what you mean. It would be interesting to attempt this
> computationally; the "transitivity" if you will of the blocking rules makes
> this seem like it could be subtler than either of us are seeing. You're
> closer to right than I was, though, if not actually right.

But the blocking rules are for the most part not transitive. if b
blocks p and p blocks f, it doesn't follow that b blocks f. The only
transitive group is k/x/g. If I'm not mistaken, you can mostly do it
with only two vowels. You only need 3 vowels when k/g/x are involved.
The blocking rules sort all the forms in groups of up to 32 members.
For example, you can sort the blanu group (16 members) with only two
vowels as follows:

blanu bramu plamu pranu flanu framu vlamu vranu
plana prama blama brana vlana vrama flama frana

For the kakne group (18 members) you need three vowels:

kakne kagme xaxme xagne gakme gaxne
kagni kaxmi xakmi xaxni gakni gagmi
kakma kaxna xakna xagma gaxma gagna

The most numerous groups contain 32 members, but can be sorted with
only two vowels.

So you could in principle have a maximum of 20315 gismu forms and
respect all the current forbidden clusters and blocking rules.

[Jorge Llambías]

On Wed, Aug 1, 2012 at 9:50 PM, Ian Johnson <blindb...@gmail.com> wrote:
>
> 5 vowels
> 17 consonants
> 48 initial clusters
> {11 unvoiced consonants, 10 voiced consonants, (m, l, n, and r get double
> counted) 9 "special exceptions": cs sc jz zj cx kx xc xk mz are all
> forbidden, and no double consonants}->191 internal clusters
> gives exactly 20315 gimtai.

I get 179 for the number of internal clusters, so the total number of
non-conflicting forms comes out as 19295.

[John E Clifford]

xorxes' note shows that the old LoCCan3 calculations are mistaken for the basic info and so probably for all the info.  But the errors are always of being too low, so the general point about every gismu having a unique form hedged in various ways remains, more strongly, perhaps.

---

From: Jorge Llambías <jjlla...@gmail.com>
To: loj...@googlegroups.com
Sent: Thursday, August 2, 2012 6:52 PM
Subject: Re: [lojban] racli/ralci

--
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+unsub...@googlegroups.com.