The Language Generator

[shinkarom]

For almost six years I've been developing a conlang generator.
It's always in rough shape. I would like very much to hear your feedback and advice .
https://mega.co.nz/#!aQ8zgJgA!feRIh6Gi_ ... SsU8ljTjdg

[opipik]

That link doesn't work.
Yes, it redirects to the server, but that's all it does and no text shows.
Because this sounds like a good idea, can you give a new link?

[ęzo]

[razed and salted]

[shinkarom]

1. I'm not an expert in programming.
2. In linguistics I'm even weaker.
3. Trust me, if I knew how to diversify grammar by now, I'd implement it by now; that's why I need your expertise.
4. The third person pronouns are "he", "she" and "it".
5. I thought that generating some kind of good phonology supercedes improving the grammar.
6. I can upload the Delphi sources and give you the link so that you look what can be done there.

[Daedolon]

Nice SAE eurolang generator

[Pole, the]

Delphi

please no

[Torco]

i second what Daedolon said: it seems to naturalize quite a bit the oddities of english. he she it has no reason to be... they're all ta1 in mando, for example, y en castellano no tenemos it y nos va muy bien. . what about tu-vous distinctions? why only animacy in the third pronoun?

Even simpler stuff like an algorythm for "okay, first choose whether animacy is marked in nouns, *then* whether it is marked only in some nouns and not others, *then* choose which have marked animacy, and *then* figure out how many forms you mean for the pronouns [depending on the answers to these and other questions, it could be anything from 1person, 2person to i you he she it, to a list of seventy-three forms where six degrees of animacy interact complexely with TAM and MSA].

I suppose some kind of iterative process would be required: choosing an overal isolang-aglulang-fusiolang typology, and so on and so forth. Grammar is a hell of a lot more complicated than phonemic inventory and phonotactics after all, and someone would need to construct a whole application in order for the relevant routines to be written. Like, sure, people could contribute stuff like "okay, i'll write a routine for tonal languages" or whatever, and that's easy, just add phonemic suprasegmental support to the thing, but "okay, i'll write a routine for polysynthetic languages" is... well, quite a bit harder: one would need some sort of formal language to describe grammar in general...

does such a thing exist ?

[hwhatting]

But if you automatize all these things, what fun in conlanging is left?
I can understand that people would want to automatize the tedious parts, like applying sound changes to derive daughter languages or to create lists of phonologically admissible words, even if I myself prefer to do these things by hand. I can also see the use of something like Gleb, creating random phonologies as a challenge. But if you use a program to automatize everything, why do you even conlang?

[Torco]

Well, for one thing, that this software exists doesn't preclude one from doing it by hand: We have snazzy 3d printers that can print sculptures but we still enjoy, appreciate, and produce wooden sculptures through the age of process oldentimes sculptors used. Printers don't make paintings any less viable, they just just take away the painting jobs which don't make much sense because they're not about artistry [getting an idea of someone's visage, for example], or not much skill beyond mechanical reproduction at all [generating hundreds and hundreds of paintings of the same thing so they can be propagated through a popultion].

But also, different people find different things tedious. I've been paid to conlang, something to do with a system of relative pronouns or somesuch a couple of years ago, because a user of this board was like "meh, I don't get this feature, I don't much care for this feature, fuck it, let someone else do this one bit of this language" the way someone might like to paint and assemble model airplanes but, say, leave the intricacies of fashioning the engine to someone else... doesn't make them any less of a tiny plane enthusiast.

Finally, automatizing makes the effort-result structure different, so conlanging with such a tool <not this one, but an ostensible very good, very advanced version of this one> might be more about trying out new things, attempting odd combinations, using strange structures and immediately seeing if they work or not, so you can tinker and tune more than lovingly craft. As it stands, to fashion a language is a laborious process... no one will live so long as to make 75 different, well developped conlangs in their lives the way a painter can paint that number of paintings in a couple of years. <or decades, i guess, depending>.

Now, whether a series of instantiations of a multivariate model can be *said* to be a language is complicated.... is the theoretical output file of this putative application a *lang*, or just a template of a lang? well, of course not... is a set of variables plugged into universe sandbox a planet? But is might be a template or a tool in conlangmaking, just like, say, RWGs or the frequentizer, or dictionary management systems are.

Come to think of it, we kind of lack a syntax equivalent of those things... you know, plug in words, POS and syntactical rules and it gives you valid sentences, or maybe it shows you the syntactic structure relevant and allows you only to put words on it in such and such way?

[Chagen]

But if you automatize all these things, what fun in conlanging is left?
I can understand that people would want to automatize the tedious parts, like applying sound changes to derive daughter languages or to create lists of phonologically admissible words, even if I myself prefer to do these things by hand. I can also see the use of something like Gleb, creating random phonologies as a challenge. But if you use a program to automatize everything, why do you even conlang?

Because procedural generation is an interesting and hot field of computing?

I for one find generated art to be fascinating and so I'm interesting in this. The beauty of raw computer code creating something from chaos is definitely different than a hand-crafted thing, but that doesn't mean it's not intriguing.

[shinkarom]

First, he, she and it were made because I like to make sentences involving females, and without a dedicated pronoun those sentences would be a lot less feminine.
Of course, I can make many changes (such as making really unpredictable grammar), but you must look at the example sentences. To make them, I needed to write a simple kind of machine translation. This translator is clumsy. Without your help it will be impossible to improve the algorithm further.
Also, I didn't say that this version is final. I posted it here just because one person on lingvoforum.net complained about me changing interface language from Russian to English.
On the third hand, to generate a random relex, even with a static grammar, isn't too insignificant a task.
The only reason I began this project was because there were no language generators whatsoever (even Gleb is too specialized). And I wasn't working on it for the whole six years, there were multi-year hiatuses.
Torco, can I recruit you to my mentors?

[hwhatting]

@ Torco & Chagen:
Yea, I get your points. I kno that the fact that something has no appeal for me doesn't mean that it has no appeal fopr anybody. For me, the process is the appeal of conlanging, not the result, but that'd just my view, so my post was more a personal vent than a challenge.

[shinkarom]

If any of you would be by chance interested in newer versions, what features you'd like to be included next? (Please tell about immediate features, not large scale changes).
It's a daring request, but more feedback is needed.

<not this one, but an ostensible very good, very advanced version of this one>

The program can't become even a little advanced without your participation.