Theiling Online    Sitemap    Conlang Mailing List HQ   

Re: Non-linear full-2d writing (again)

From:Sai Emrys <sai@...>
Date:Monday, January 23, 2006, 6:01
> I've picked up Harpelan again, a little bit, and I'm slowly refurbishing > the morphology and lexicon. (I'm still nomail.)
... yet somehow you saw this. Magic! :-P
> > * how to have sememes that overlap or intertwine? > > Sort of a case study: I haven't done much of this in Harpelan, with the > exception of some outside/inside pairings. In my case, I've found the > results aesthetically unpleasing. The exception is Harpelan verbs, which > comprise a small and closed set (currently numbering seven) that take > affixes to mark modality, intent, and auxiliary verbs. All the verbs have > approximately a square or pentagonal shape and have a large blank space in > the middle -- this gives them reference points so that sememes can be lined > up consistently; even so, there's quite a few not-quite-fits that become > special cases. In all, it's hard to make characters that are visually > distinct on their own, yet have a common set of reference points to align > with other characters.
Mm. What if your verbs were not all of the same shape - in such a manner that they would overlap well? Or for that matter, you could only have overlapping with things that would make sense as modifying each other - e.g. integrating object and subject into a verb; perhaps its modifiers as well (adverbials etc)...
> It's also a challenge to make it so that the individual sememes are still > recognizable after overlapping. Currently, I don't see individual sememes > in verbs. Instead, I see first the density of features, which tells me how > many sememes are there, and then specific distinguishing features such as > "equals sign" double lines that are characteristic of particular sememes.
Well, they needn't be individually recognizable so much as their mesh needs to be comprehensible. If it can be comprehended with enough ease in its combined form, I don't see any need to worry about it being difficult to pick apart its subparts (any more than with 'normal' fusional morphology...) I'll be honest though and say I currently have no idea how to make that happen. I think it can be done, but that's mostly faith.
> > * how to have the 'web' interconnected at higher-order levels (e.g. > > paragraph plus)? > > That's what names are for =) > > There's a graphical programming language called "Labview" where, instead of > typing programs, you draw them by laying down boxes (which represent > function calls or structured control flow), and connecting the outputs of > boxes to the inputs of other boxes. Labview has this wire clutter problem: > for modest to large size programs, one ends up spending more and more time > adjusting the wires that connect different boxes together so that the > program is readable. I think this is kind of an intrinsic problem with > graphical writing; if you want to connect related things together, you're > going to get wire clutter as the text gets larger. The solution is to name > things that are not local to a small area. I'd like to hear any solutions > others come up with.
Agreed. Of course, it assumes a particular kind of connection display (direct point-to-point lines). Perhaps you could have things that *point*, or some sort of automatic naming (e.g. a hash, or a much-simplified version of the thing referenced), or ...? My point though was just getting it away from capping out at chunked sentences to start with. It seems you've got that one down and are dealing with how to resolve clutter once you have a complicated structure. I would think that you could use a computer to help with this - e.g. one of those programs that makes nice little graphs of social networks, and puts weights / repulsions on them all so that you have relatively little overlapping. But that's suboptimal, if you want to be able to write it normally. The other aspect of this is aesthetic. I find it a bit jarring to have a nice graph / web structure... and then see another, totally separate, next to it. Back to linearity again, just with a higher chunk level. Surely there's some way to have it pretty & comprehensible, and still look organic? - Sai