Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[irmo]

It bothers me that the dwarven language used for names and such is hardcoded rather than procedurally generated.

Before worldgen, there needs to be a "pre-worldgen" phase where the game generates things such as pronounceable roots and affixes for words (and also creature anatomy, but one thing at a time). It should then divide these up using a clustering algorithm to decide which consonant groups belong in each language, and then assign each one a system for representing parts of speech and verb tenses (such as isolative, agglomerative, inflected, etc.). Since this happens before worldgen, it can affect things like the initial name of the world and the user interface (like instead of the menu saying "Dwarf Fortress", it will have the dwarven words for "dwarf" and "fortress").

Then, during worldgen, the first time a dwarf encounters some object or activity, the language table fills in that slot with a randomly chosen word. This table is used to name dwarves and their settlements and artifacts, so very early dwarves all have names that mean things like "stone" and "beard", and later get names having to do with pottery and metalwork and such.

[FallingWhale]

For dwarves no; for animal men it's planned.

[Niyazov]

The world of DF does not have a prehistory. As with western man's perception of the world until about the 18th century or so, the worlds of DF are created out of nothing in a short time and in just about their current form, and their brief histories are fully documented. The idea that there might be periods of history that were not historically recorded (in sacred books at first, transitioning into less trustworthy pagan histories around the time of the Greeks) has only gained currency recently. This belief still endures in many places in the USA, where young earth creationists will happily explain to you how the world was created 6000 years ago and are hard at work to introduce this idea into public school curricula.

The fantasy cosmology of DF has its roots in medieval thinking; it is haunted by ghosts and monsters, vermin arise as if by magic from the earth, and its inhabitants live and die according to regular cycles without thought of advancing their knowledge or station. Medieval scholars were of the opinion that the divine language spoken by Adam and his kin up until Noah had existed in a perfect form, with words for everything under the sun, even those which had not been brought into existence yet. Medieval religious ceremonies were conducted in a dessicated and unchanging Latin that had preserved its form long after its function had been exhausted; what education there was was conducted in Greek and Hebrew that would have been familiar to speakers of those languages a thousand years before.

It makes sense that a fantasy world brought forth out of nothing with a history of scant hundreds of years, uninterrupted by deluge or Babel, should have unchanged languages retained since the beginning. Elves and goblins can be met in any world who have no parents, as they were brought into existence from nothing when the world was formed. Why should their languages change? Few men or dwarves can claim a lineage beyond a few tens of generations. For them there was no long journey to civilization, no loose tribes united by common need, no great migrations or new discoveries. No proto-dwarf ancestor ever crawled from a limpid cave pool to lie gasping on a dark rock at the root of a mountain; no ape-man ever snatched a branch from a burning tree after a lightning strike on a dark night and conquered fire; the secret of making steel and forging armor and weapons were not discovered but have always existed from the very beginning.

The world of dwarf fortress is very different than our own. This is also true of its languages. The broad sweep of time that we think of when considering human history simply does not exist.

[Sadrice]

Stirring though your post may be, I suspect that's less a stylistic choice on toady's part, and more a matter of him not having gotten around to implenting it yet.

[rhesusmacabre]

Toady's expressed interest in doing more with languages, but has also mentioned the potential problems (from DF Talk#10):

The other issue is if you're having the game randomly generate a language, which is something that we've avoided up to this point, even though you can throw something together, especially with as little structure as our languages have now – we don't really have any grammar for them or anything, it's just here's a word, here's how you write it – the only reason we haven't done that is because we're afraid of getting too many words that are real life words. So for our dwarf language we didn't want the word for 'mine' to be randomly generated as 'pepsi', and have the word for 'dwarf' just be some swear word because of what it comes up with ... so if your name is 'Dwarf Mine' your name would be like 'Poop Pepsi' or something. It's just not what we imagined for the dwarf language.

[Babylon]

My main problem with this idea is that currently the language files are editable and if they were procedurally generated they would not be.  I like the option of having the Dwarves use German, or Klingon, or whatever.

[GreatWyrmGold]

You know what else bugs me? The text for the game is always English! The game should randomly generate which language the interface is in!
And you know another great thing about this suggestion? It would turn the whole namng system of "Urist McDwarf" into an anachronism! It's almost as good as if we were forced to trade with elves instead of killing them!

In case you can't tell, that was sarcasm. Seriously, some things can remain constant between worlds.

[Starver]

In many ways, this suggestion reminds me of one of the earliest threads I participated in... ah, here we are...

http://www.bay12forums.com/smf/index.php?topic=38052.0

Now, it wasn't really talking so much about how the game conveys itself to the user, but within some of the ideas discussed are the possibilities that "token terms" and a structural matrix to convey these terms could be procedurally assigned "output tokens" relevant to the original terms (and given a racial flavour), and with less reliance upon strict "subject object verb"/whatever formation, but leave that to a procedural (tightly-seeded PRNG originating) creation process suited towards the target as well.


OTOH, I think we'd have to constrain the process a lot (probably to remove most of the variety that could be experience to avoid the risk of Dwarves having 'moonspeak' or even 'Clanger' words; or risk harsh linguistic constructions that even a half-Welsh half-Klingon with a bad cold, irritable mood and particularly tight collar would rarely utter.  What we've got right now does, to me, sound about right.  But then it would, I suppose, being what I'm used to.

[Jeoshua]

Here's why random words shouldn't be totally random:

Dwarf named "Kriusoanotuyhi McNeuoauntaturioal"

Totally valid linguistically, could only really be generated randomly, almost completely incomprehensible, and not nearly Dwarfy at all.

[Reudh]

What you could do is use the utility DFLang and make your own generated language for each world. Only a small effort on your part and it does exactly what you want.

[Courtesy Arloban]

What you could do is use the utility DFLang and make your own generated language for each world. Only a small effort on your part and it does exactly what you want.

If you put in some more effort after using DFLang to switch the generated words around, you can have a more natural sounding language.  I got the idea from a program made for linguists that did the same thing as DFLang, but generated a spreadsheet.

Here's why random words shouldn't be totally random:

Dwarf named "Kriusoanotuyhi McNeuoauntaturioal"

Totally valid linguistically, could only really be generated randomly, almost completely incomprehensible, and not nearly Dwarfy at all.

You can avoid that problem  by having a foreign dictonary and find the procedurely generated word that comes closest to the translation for the english word,  you avoid it being too close of a match by generating the raws with a different word list.  You can convert raws, including generated raws to a spreadsheet using search and replace in a text editor, if it makes it easier for you, by replacing double quotes to single quotes, and brackets to double quotes.

The program was updated to no longer use the spreadsheet because of licensing, before it was discontinued.

[Escapism]

You know what else bugs me? The text for the game is always English! The game should randomly generate which language the interface is in!
And you know another great thing about this suggestion? It would turn the whole namng system of "Urist McDwarf" into an anachronism! It's almost as good as if we were forced to trade with elves instead of killing them!

In case you can't tell, that was sarcasm. Seriously, some things can remain constant between worlds.

Pssh, you need to start thinking on a greater scale. Randomly generated UI language? What we need, gentlemen, is a procedurally generated game.

[Starver]

Pssh, you need to start thinking on a greater scale. Randomly generated UI language? What we need, gentlemen, is a procedurally generated game.

I won't go into the details (because it's nowhere near ready for the public, anyway), but I have something like that in pre-pre-pre-development.

Spoiler: But that's off-topic (click to show/hide)

It's currently called "Outcrop", should any future historians (or historic futurists?) have any reason to look for a bit of basic documentation about it.  The start sequence is deliberately ambiguous and when you actually start playing the game, at the spot that your "potted world overview" has zoomed into (the same spot, with the exact same things dotted around!) you then play as someone apparently with amnesia (or... you might never have known certain things... but you as a player do not know that... yet, or perhaps ever!) and then you make your way around the environment and either work out what your situation is or just "do your own thing".

I had rather imagined that each and every separate installation of the game would procedurally derive the many, many hidden constants (or generate them as and when they were needed) from the state of the system (HDD serial number, free space on drive, current free bytes of memory) in a non-linear manner so that each and every install[1] is different so that one person might be playing in a virtual war-zone (until and unless they can engineer peace, either through personal force of arms or some more socially-interactive manner) or else something no more strife-torn than the average The Sims neighbourhood (writ large... at least until and unless the player sends the place into war by random acts of terror).

I don't know how many people would actively 'buy into' that.  I've a feeling there'd be a big market in 'patches' to turn your "get on with people, earn money, get a bigger house"-style game that some people would find themselves having installed into the urban-combat simulator that it could have been, but wasn't.  I'm rather of a mind to make it highly obfuscated behind the scenes to make it very much less than trivial to either predict or influence the game in such ways.  And if your first action upon 'waking' is to go and find the next random stranger and break their neck...  well, maybe the dragons will take over the world.  (Dragons?  Well, in this instance of the game there are dragons.  Whereas in another there may be aliens and a third has an evil-genius-in-an-extinct-volcano.  A fourth install's biggest dangers is some minor crop failure due to drought on the eastern coast.  A fifth one, meanwhile, is in direct danger from you for various reasons that I won't go into right now (and when you realise this do you fulfil this danger or work to prevent it, eh?)).


But it could go further.  I currently have the game as a 1st-person (or over-the-shoulder 3rd-person) game, in a fully rendered vector-based landscape upon a pre-seeded (i.e. static) procedurally-generated landscape with key 'deltas' being applied on top to ensure various features (townscapes, infrastructure) appear in locales where they don't always arrive...  But, first of all, if I was more confident that my procedural generation would still create legitimate landscapes (without 'rejects and resets', a la DF, hopefully) maybe I could give each and every person a completely different world...  And to be truly free, I could make the interface much looser.  Only a few small changes to the rendering engine could change it from ultra-realistic 3D to 'cartoon' representation, and it could as easily be represented in DF-style, or textual adventure mode.



[1] Actually, I quite like the idea of making it so that it is the same game on the same PC, at least until you change your HDD/hardware and reinstall.  Or someone 'tricks' the program into changing (blindly, or having worked out how to aim at a target scenario), but I'm not sure whether the market would bear this feature.  And imagine the game reviewers raving over "The best social simulator I've ever played", and some buyers finding they've got a high-fantasy game installed where they're needing to forge a better magic-proof sword, while others find themselves leading a forces of guerillas in urban house-to-house battles in an attempt to overthrow a corrupt politician...

[Lipbalm]

Now, if you're after a realistic Dwarven language... I dunno. It'll be tricky.

Firstly, is there a need? Yeah, almost everything's randomly generated but procedurally generating an entire language is pushing the limits a bit too far...

Let me explain in this handy little spoiler:

Spoiler (click to show/hide)

Vowels and Consonants are easy... I can see that happening.

Grammar: Here your problems lie.
Realistic languages change over time so the language generated at year 1 may be different to year 1051. How different? Look at Latin and the Romance languages. The language may have to change depending on the migration pattern of the Dwarves, Elves, Humans and Goblins and then you have to account for the presence of certain items or lack of (e.g., One civ may not have lobsters and another will but they speak different languages. How to they swap words? Do they assimilate the phonotactics or adjust it accordingly?)

There are also 6 possible word orders (syntax):
SVO - I boarded the ship
SOV - I the ship boarded
VSO - Boarded I the ship
VOS - Boarded the ship I
OSV - The ship I boarded
OVS - The ship boarded I

Some languages even ignore word order and rely on cases. How many cases are there? The wikipedia page lists over 100 cases.

Do you have postpositions or prepositions?
I am on the ship -- Preposition
I am the ship on -- Postposition

Do you have words like "the" and "a(n)"?

Are you allowed to drop personal pronouns?

How does the languages verb structure work? Do the aspects, tense markers, etc just attach to each other like Navajo does or do they work like English or Chinese?

... And many more!

All of these must be answered for a conlang. Which is what this is. Some historical stuff can't be emulated in a game -- even Dwarf Fortress -- just because it'd take too long to generate.

In any case, all I can see is the languages of each civilisation at completely different ends of the word being similar or exactly the same. This happened with the Polynesian languages (Maori, Samoan, etc) and they're really similar because they've only had less then 2,000 years to separate. You're not going to end up with a language situation like Europe in Dwarf Fortress. It'd be nice... But unless it exports the entire grammar to a PDF I'm not liking the idea at all. It'd take too long to generate a world.

What I'd rather is if the game picked out one of several templates for languages which could be applied at generation. It's a good idea but... It's only a good idea in theory. However, if there's an easy way for this to happen and not make me cry whenever I see the language, I'd be all for it.

[Nonsequitorian]

Because of Dwarven Efficiency, and because it's hinted at in the language raws, I imagine that Dwarves use cases, adjectives are just tacked onto nouns, and word placement is the same as Latin: not usually relevant.

I personally love it being like that. Having done a whole lot of Latin in my years of schooling, and having been really very good at it, I find the case system and lack of word order (for the most part) to be really interesting, if slow, and that's what the dwarven race needs. Lets say a dwarf wants to say "I want a red apple," I see him saying something like (and I'm just making this up on the fly) "Mœlar urd bemīngrang"

Mœl would mean "To Like" (because I couldn't find what it actually is in DF). Ar may be the ending for present-tense verbs

"urd" would be "an/a."

bemīng is apple in the nominative, bemīnger is the accusative, but bemīngr has no "e" because ang (red) starts with a vowel.

Another example:

Spoiler (click to show/hide)

"He thinks I am hateful" could be translated as "Tholim ner Summl lurnter."
Thol is think, as zotthol is unthinkable. I imagine regular verbs ending in the letter L just because why the heck not. "Im" may be the ending for present-active 3rd person singular, so Tholim means "he thinks."

Ner could just be the accusative form of "I." In English this is usually the word me, but not in this case. It is the word me in latin, however, and because there is no "that" in latin (not as in "that car," but as in "I think that blablabla"), I don't see one here.

Summl is the infinitive of the english verb "to be," just because sum is the latin and I want to have at least a few slightly connectable words.

Lurnter comes from the dwarven word lur, meaning hate. To hate could be as simple as adding an l to the end, so lur "hate" becomes lurl "to hate". As I see it, hateful and hating (the gerund form) mean the same thing, and to signify that lurnter is both gerund and accusative (accusative for the sole reason that ner is accusative, so we can see that I am hateful, not he who thinks is hateful), there is a gerund ending (nt) plus an accusative ending (er). Why not Lurlnter? Try saying that ten times fast. Heck try saying it once! Gerunds should just drop the l and add nt.

Personally, as a lover of language, I find the that quirks add flavor to what would otherwise just be bland and boring. Just for the sake of it, I think verbs should be always capitalized. Thol is capitalized, Summl, Mœl too.

Speaking of which, if you understand at all what I'm talking about then I do believe we have enough people to start a project.

I think that we of Dwarf Fortress should edify the dwarven language in all its grammar and glory. Words would have to be created, but if ten people work on it, then adding 100 words will be easy. Anybody think it's as good an idea as I do?