Re: Non-linear / full-2d writing systems?

... and back into the fray. Now I'm all read up on this thread again,
though I'll probably find another halfdozen long replies by the time I
am settled in DC with an operational computer... (Not that this would
be a bad thing at all. :-))


Ray Brown:
---
But by 'more atomic', I was not intending to refer to 'atoms of meaning'
but to 'atoms of the thought process' (if that makes sense). Just as
speech - the primary medium of language for the past 170 millennia or so
of Homo Sapiens*, is analyzable into phones, I just wondered if thought
was analyzable into 'basic units of thinking' - probably not, I guess, but
Cognitive Science s very much a terra_incognita to me.
---

*laugh* If you have any idea what they are, please let me know.

I know the NTL group's idea of it, and Lakoff's rather handwavy
metaphor-based one, but they all feel rather incomplete to me. I'm not
certain whether they could be applied to actual language-building, and
howso if yes. I'll percolate on it.


---
What goes on in the brain during thought? I suppose what I was thinking of
was something that would map the brain activity, so that the reader would
then be able to recreate the same activity in his/her brain. But I am now
probably entering the realm of science fiction  :P
---

AFAIK that'd be impossible simply because of the order of magnitude of
activity going on. Seriously, do you have any idea how many neurons
you have, with the branching factor and number of (asynchronous)
messages per second of each? Oy gevalt. And then, figuring out just
what subset of that constitutes "a thought" (read: consciousness) is
an extremely open question.

And yeah, you have the problem of translation; at that low a level, it
would be rather silly to try to map it onto someone else's brain; you
do need to have some level of abstraction. That's where the trouble
comes in. :-P

(BTW, Lakoff would just love the opportunity to [yet again] tear up
Platonic categories. Read "Women, Fire, and Dangerous Things" if you
want his perspective on that.)


---
> However, I think it's a good metric for how
> worthwhile the 2d-ness is; the more irrecoverable the translation, the
> more added value it is giving you, QED.
Absolutely - if there is no added value then there seems to me no point in
making the 2d thing, except maybe as a fun thing.
---

I think this is where the distinction is, for trying to explain it to
non-initiates; might be better to emphasize this in trying to convey
the idea. I think the initial skepticism is partially that this is
"just another form", and hence merely an aesthetic thing - i.e. not
worth the effort to learn. But if we can at least suggest - and
hopefully, at some point, implement - value-added features, it would
be a stronger argument. After all, all you need to proove is that a
NLF2DWS can be at least *somewhat* better than any given L2DWS; after
that, it's just a matter of "bargaining over the price". Induction.


---
> But I would point to my earlier mention of "translation" as opposed to
> "interpretation" etc. That is, you would need to recover the *actual*
> meaning of the text - background and all, if necessary - and try to
> write *that*.
Not easy. In fact IME all translation, except of fairly trivial stuff,
involves some loss and some degree of interpretation - traduttore
traditore.
---

*chuckle* Yeah. How do you translate puns? Metaphor? Etc... it's a pain.

Which is why ASL interpreters are rather resistant to the word "translator". ;-)


---
> I think a very related thing that would be necessary to research (and
> I know that a lot of research has already been done on it, it's just
> not stuff I've read) is on visual design and presentation in general.
So much to research! If i had nine lifetimes, like the proverbial cat, I
might manage something  :)
---

*laugh* Likewise. Ah, well, I'll have to make do with my 45 year
estimated productive lifespan...



Schuyler:
---
I'd like to introduce myself to this list.  Hello! Sai directed me to this
community due to a post about my conlang (with 2d writing) on livejournal:
---

Welcome to the Other little list. :-P


---
1. Anything finitely representable (including anything with finite inputs)
can be represented by a 1-dimensional string.  There's no use making a
definition that says something is irreducible to a 1d string.  However,
we can make a weaker requirement that our 2d communication cannot morph
into a topological 1-d thread.  That is, there must be at least one loop
(where ends reconnect) or one branch.  From this, for it to be
interesting, loops and/or branches must have additional communicative
value in the 2d writing system.
---

Agreed. Is it attackable on the "finitely" side also?


---
2. With that, I'd like to propose a more general definition of 2d-writing
that turns some of the past debate in this thread into a variable of the
writing system:  The loops and branches can be most common in a writing
system at either:
   a. the word level (including iconographic languages,
       like cuneiform and Chinese)
   b. the phrase level (a la sentence diagramming)
OR
   c. the sentence/multi-phrase level.
---

Hm. "Loops and branches" at a word level? I don't understand how
you're applying that.

Also, you're missing another thing that 2d brings: more positioning.
Linear you only have "before" and "after", + distance (maybe); 2d you
have 360 degrees + distance (less if broken down into something
distinguishable, but at least 9 or so).

You also have potential orientational meaning - i.e. VX might be
different from >X or <X (assume V/</> is the same character). This
would be permuatative.


---
2. Personally, I'm interested in human-usable 2d writing, which adds a
host of constraints.  With that, I think 2d writing can be interesting
even if it doesn't do something unique that 1d writing can also do.  It
just needs to do something *better*.
----

*nod* Which is what I meant when I was talking about the NL2d->L2d
translation being too lossy. That is, it's almost certainly possible
at least on a theoretical level... but so is programming a major
project in byte code. :-P


---
3. Some things that I think 2d writing can do better than 1d:
   a. comparing abstract concepts or arguments to each other.
   b. Since computers have 2d screens, 2d writing is a perfect opportunity
to add computer-context additions to the writing system.  Generally, this
relates to the easy mutability of computer texts--so change-logs, and
identifying many authors in a threaded text (like conlang-l archives, for
instance :-)
   c. Giving a text map-like qualities--this is something that outline
form does for linear writing systems.  2d writing can give us the
opportunity to 'zoom in' and 'zoom out'
---

*nodnod* Though you run into the query of what constitutes a proper -
or (a rather different thing) a *convincing* - argument. (Humans are
really not logical machines, no matter what some folk wist for...
[that, and I can hear Lakoff wanting to interject how "logic" is
actually a metaphor for some handwavy 'embodied experience'].)

For maps: What makes for a good map? That is, one that gives the most
information at first glance, on search, on detail, etc. Colors are I
think one major thing, as well as size of text / line etc. One problem
that I think will need to be dealt with (and I see in, e.g., your
constitution example) is having too much being processed at once, and
being overwhelming. You would need either something that is built to
handle "spotlighting" - that is, being read in some particular order
and intentionally ignoring the rest to do that focus - or to handle
chunking (like a map), where you basically have a cognitive 'zoom',
and any particular level of zoom/focus only has a graspable (7+-2?)
amount of information-chunks.

Another qualification would be that of databases: difficulty of
browsing, search, retrieval, insertion, and deletion.


---
   d. Related to (c) is that related ideas have another axis to nuzzle up
to each other.
---

Muh?


---
   e. 2d gives more freedom to the reader in what order they want to read
the text.  1d largely gives a single order.  2d gives at least as many
ways as there are loops and branches (and more likely 2^n ways).
---

*nod* So do you have ideas for how to deal with the "but I want my
linearity!" complaint (in re. stories, etc.)? For that matter,
order-of-presentation has been a pretty important part of writing for
a while now - e.g. for presentations, arguments, stories, instruction,
etc. Presumably a web could be made to be more easily traversable in
particular ways, due to the arrangement - in fact, I suspect it would
be rather difficult to *avoid* having bias on reading order - so we'd
need a fairly different way to handle it.

Not to mention the more subtle technique of trying to play with
order-of-comprehension - what I think Teoh called the "aha!" form of
storytelling. ;-)


---
My 2d writing system was developed somewhat before I knew I wanted to do
all these things, so I'm still working on including them in a way that is
satisfiably integrated.
---

*laugh* Sounds like my entire conlanging career. But I got derailed
entirely into working on figuring out wtf I wanted to do on a meta /
theoretical level...



Ray Brown again:
---
When Sai wrote "not linearizable without loosing damn near everything in
the process in a way that's cognitively irrecoverable" I did not
understand him to mean that it could not be put into a string with
appropriate parentheses & other symbols without those losses. I understood
in the context of this thread to mean that it could not be put into a form
suitable for 'ordinary' spoken language.
---

Pretty much. The key part is "cognitively irrecoverable". Like your
tree vs paren example - some formats are fundamentally a whole lot
easier for us humans to understand. Pattern-matching is a lot of it,
IMO, as is how well it can chunk, and how well it can be integrated /
related to other things we already know. (E.g. my rote memory is
horrible, but I learn things that I can make sense of and integrate as
worldview / how-the-world-works extremely well.)

So in that sense, while format may be irrelevant in the formal sense
(a la Turing et al), it sure as hell is important for *us*. That's
where the lossiness comes in - not from loss of *content* (I'll agree
that anything can most likely be arbitrarily encoded into any form,
including 1d), but of *comprehension*.


(Side note: How many people reading / talking about this are CS/math
geeks, or at least enough to know about digraphs, arrays, stacks, and
Turing machines?)


---
What I understand is this:
- the linear writing of the past 5 millennia has been recording of spoken
(or speakable) language; as the latter is obviously linear, therefore the
representation of it has been so.
- many people do not think in words or do not think entirely words and
find that 'words get in the way'; they find there is inevitable loss when
words have to be use.
- Sai suggested that a non-linear and fully 2d writing system might be
more appropriate for representing thoughts without 'words getting in the
way'.
---

*nodnod*


---
But this is not representing the 'thought behind the words'.

OK - yes, I know you are wanting to broaden the scope of this thread. But
(b) & (c) is significantly, I think, different from what I thought was the
original intent of the thread. I could well come up with a scheme for
(b) & (c) by the June 30th deadline - but not for what I understand was
the  sort of writing system Sai had in mind. That is rather more difficult.
---

Hmm.

For me, the two are separate but very interrelated.

For one, there is the question of form, as Sky pointed out. I.e.
ubiquitous 2d-value-added (hence, loops, branches, & positioning).
This in itself is IMO a worthy endeavor, since it can by itself be a
good argument for NLF2d being "worth it".

For two, there is the question of content. Or more accurately, of
"nativeness" / "intuitiveness", as I might call it - that is, how
close is the manifestation to the internal source? We all start with
something underlying a "speech act" (heh) - some communicative intent,
or a thought, etc. I would like the manifestation of that to be as
close as possible to the "real thing".

This is because I make a couple assumptions I should probably make explicit:
a) Closeness -> ease of understanding, if you already share a similar
background. IOW, it ought to be more intuitive, more easily
understood.
b) Closeness -> ease of expression (once you're used to it and perhaps
"if sufficiently introspective")
c) Closeness -> minimization of translation error
d) There *is* an underlying, nonlinguistic, thought.
e) Unintentional misunderstanding is the primary cause of inter-entity conflict.

(I'm probably not wording this well and leaving out a lot; it's a
first-pass attempt.)

So for this latter part, yes, telepathy would be the ideal case. (Of
course, you run into problems when trying to communicate with someone
when the differences of your backgrounds come up... but this is
already a problem [political debate, anyone?], and at least it would
be more explicit.)

A simple example of intuitiveness: IMHO, ASL is *WAY* better than
English at describing events that involve spatial movement. E.g.
describing an automobile accident, or how someone was walking... it's
denser and more eloquent than words, inherently. Can that be extended
to *thought*, though? Or to everything?

(BTW, Sky: What "constraints" do you think human cognition imposes on
2d language?)


Sky again:
---
If anything, a 2d writing system will 'get in the way' more so than a 1d
writing system, because implicitly a 2d writing system pre-defines the
axes in some way.  That is, whether 2d writing is defined by axis (Remi's
example from May 8th), intersection (like mine), branch, loop or some
combination of these, definitions will make 'free' use of the 2d space
less possible, not more.
---

*nod* True. Depending on the form, of course; I think axis- (Remi's)
or fixed-positional- (Pinuyo's) based use would be a lot more
restrictive than yours, or of atom-and-relation. But then, does it
matter? I don't believe that there *exists* so-called "free" use of
anything linguistic (including "free word order"); it, too, conveys
something - generally subtle shades of pragmatic meaning or emotional
effect.

So IMO, it is utterly useless to have "free" anything that can be
utilized grammatically... but I am very much for having emotional
effect etc. being treated as equally worthwhile aspects of grammar. (I
find it rather perplexing when a linguist tells me that some
particular phrase is in "free variation" with another, and handwaves
the usage differences. WTF, man? I thought linguistics was about
figuring out how we use language; why pretend that inflection doesn't
exist just 'cause it's hard to pin down?)


---
Before representing the 'thought behind the words' you need to know what
kind of structures (nouns, verbs, etc) 'pure thought' operates in.
---

Agreed. But do you really think it operates in something resembling
that sort of category?

---
I read Sai's "lines" as non-linear verbs ('is equal to', 'is in', 'is
like', etc.)
---

*nod* Sorta. Heavy influence from _Describing Morphosyntax_ on that one.


---
By (b) & (c) I really mean to describe 2d writing systems'
'resolution'--are the 'words' themselves created by 2d relations
(axes/branches/loops) or is the 2d-ness at the phrase level (e.g.
connecting adjectives and verbs to nouns; i.e. (b)), OR is the 2d-ness
connecting complete 'events' (e.g. axes represent time and space,
perhaps).
---

.... huh?, again. Could you elaborate on this and rephrase it?



Ray once more:
---
Thanks - but "my intuitive" definition is nothing more or less than my
interpretation of what Sai had in mind when he started this thread.
---

*laugh* Don't feel a need to hold to my original concept; if you can
think of (interesting) ways to extend / morph it beyond its intent, by
all means please do. I'm sure you have a different conception of the
Nature of Thought (tm) than I anyway. :-P


---
Ah - I think we may be using 'linear' differently. That's the trouble with
polysemantic natlang words.
---

*chuckle* FWIW, (somewhat OT) - I think polysemanticism is a good
thing. Compression, extensibility, conservation of space, and openness
to amusing semantic connections (like puns [yes, I just expressed
support for the existence of puns; you'll live]).


---
I have understood us to be using 'linear' in this context to mean a
'serial stream', i.e. words in a spoken language following one another. In
the sense I believe you are using 'linear' Cicero's long periods of
carefully crafted co-ordinate & subordinate clauses are not linear; but in
the sense I have been using 'linear' they are linear in that the words
must be declaimed serially.
---

*nod* I agree. Linearity of content, vs. of form. If the content is
linear, then this whole discussion becomes moot. (Unless someone can
see a benefit to representing *linear* content *nonlinearly*? I don't,
offhand...)


---
> but so far, this is how I see it:  If writing relies on
> common signifiers between writer and reader, then the only thing that gets
> in the way more than words is NOT having words.
Yet 'word' is one of those concepts that are tricky to pin down.
---

I think what you're both agreeing on is the need for common context. I
suspect that it is impossible to communicate without one, and that the
amount of overlap will strongly limit how much it is *possible* to
communicate - after all, you're only building from it.

Sky brings up the *symbolization* question... which is part of what
this is all about, ne? So long as you have mutually-agreed
symbolizations (whether into a word or a particular 2d grammatical
relation) it should be okay.

I found that _Describing Morphosyntax_ generated a bunch of ideas for
me in this vein (and wrt lexical categories, "verbs", etc) - because
it broke down in part what the form was by what it *communicated*.
E.g. the chapter on copulas: subset, equation, and descriptive
relations are all described by the verb "to be" in English... and it
would be rather misleading, IMO, to try to then translate them as if
the item in question is "doing something". They're just relations
between two ideas, and that is what should be conveyed.

Of course, it's entirely possible that there are better analyses of
this; can anyone suggest one? I haven't read much formal
linguistics...


---
Smiley's are interesting in that they developed to fulfill a need to
supply missing _non-verbal_ information.
---

Non-lexical, perhaps? I think it's certainly in the language-as-spoken...

And I'd assert that this is also a critical part of the intended
communication, and as such the language should be designed to support
it, rather than handwaving it away into "free use" or other such
nonsense. "Body language" is a part of the in-person communication,
and a grammatical one at that... I would personally rather like to see
it be taught in schools alongside verbs and adjectives. In any case,
if the communication has that - what we might call in English
"intonation" or "inflection" or other non-explicit but still
grammatical and included - it should be represented in the form,
whichever mode it's in.

Which brings up the question: how do you represent the various
emotional inflections in NL2d? And how can you make explicit grammar
for something that is implicit?


---
Maybe Sai could point us to current thinking on this in cognitive science
    :)
---

OK, a (very) short summary of the ideas given to me at Berkeley. Ask
me for more links, or better, google "berkeley" and the class number
and you'll get to the syllabus, links, etc.

* Lakoff & metaphor: (Read any of his innumerable books on the subject
for more, e.g. _Metaphors We Live By_ and _Where Mathematics Comes
From_.) - CogSci 101

Most cognition derives from metaphorical thinking based on a small
number of core, "emobided", concepts, and combinations or extensions
thereof. Doesn't really address the question of atoms; assumes though
that all relations, interactions, descriptions of causality, etc etc
etc, are metaphoric, metonymic, or image-schematic. (Close enough.)

* Neural Theory of Language (google it for the group propoganda
website) - CS 182 (class website has lots of info)

Various theories.

Assumes that all language / cognition is at least in theory reducible
to neural events; particular 'atoms' are sets of neurons, and
relations are (maybe?) links between them, or particular patterns of
firing, or ... (called "the binding problem"; unsolved).

From there, it builds up into frame semantics and simulation semantics
- i.e. to understand any phrase, you simulate it in your brain
(possibly using the mirror neuron system?) on a very literal level,
down to motor neuron activation. (Suppressed somehow by something that
puts you into simulation vs. action mode.)


... that's basically it so far. I'm sure that Seale will have yet
another perspective on how he's right and everyone else is wrong...


---
My understanding of Sai's ideas depends not solely on his opening mail,
but on other replies as well. I do not claim, of course, that my
understanding is entirely correct and I may have overstated some aspects &
perhaps understated others.
---

See above. :-P


---
"You can do that, but it's rather unusual. Most would consider e.g. maths
or formal logics a notational system, but not a writing system. Writing
systems are usually considered the subgroup of notational systems that
represent languages."

To which Sai replied:
"But we're in the business of *creative* linguistics, are we not? I for
one am not interested in constraining what language can do merely because
it hasn't been done before in a natural language."

Sai's reply struck a chord with me.
---

*laugh* That would be my Discordian side showing again. ;-)

I think the world needs more constructive chaos. In general, I am not
interested in doing something because it has been done, or because
nobody knows how to do it otherwise... but in doing it how it would be
*best*. (I'm not very good at following rules, as you may guess...)
When this is an unkown, you have to balance your judgement of how
likely it is you'll find a better way, vs. how much of a difference
that would be, vs. how much effort that would be, vs. the likelihood
that the tradition is there 'cause everyone else who tried to do it
different got killed. (Instead of just 'cause your
great-great-great-grandmother only had a single, short baking dish and
hence always chopped the sides off both sides of her meatloafs before
putting 'em in the oven...)

In this case, I am daring enough to imagine a different way of doing
things, and optimistic enough to believe that it would be better. I
have as yet no proof save that intuition.

But, honestly, folks: wtf are you doing this for? To make yet another
silly Euroclone? Why bother?

I see only two ways you can approach conlanging: as utility, or as art.

The utility of creating something that has no (or minimal) feature-set
difference from something already existing is minimal - either
encryption through obscurity, or the nice glow of having done it
*yourself damn it* and having it be yours. Or, perhaps, for the
learning process of having done it, even if the end product is a
throwaway. So, otherwise, you would want to go for new and better
features, better implementations, etc - the standard software model of
competition.

If you're going to treat it as *art*... well, I suppose that's a
rather contentious topic. But there are only two things that strike me
as art. The first is that which sublimely renders the real -
Michelangelo and the like - as a manifestation of the skill of the
artist and their perspective on the world. The second is that which
causes me to think in a new and different way - not merely by saying
"hey, look, it's a toilet" (I can do that myself, thanks), but by
actually presenting an idea I *haven't* had before, or a perspective I
haven't had before, or something else that expands my cognitive
horizons.

The latter happens rather rarely, but is something I value quite a lot
when it does.

Either way, as I figure it, there is relatively little use to
yourself, and none to others, in being a linguistic sheep when
designing a new language. (Or a new car, for that matter.)

This is why I have spent so long away from actually making a language:
I simply don't consider it worth my time and effort to do something
that isn't a better design.

I guess I'm just not cut out to accept "'cause that's how it is" as an
answer for why I should or shouldn't do something.

And as a result, I go walking around in funny clothes (I wonder what
the airport screeners will think of my greatkilt...) and do silly
things in public. ;-)


(Maybe I should write this up as "The Discordian Conlanger's Manifesto"?)


 - Sai

Replies