Re: ANNOUNCE: My new conlang S11
|From:||Henrik Theiling <theiling@...>|
|Date:||Thursday, March 3, 2005, 21:45|
"H. S. Teoh" <hsteoh@...> writes:
> On Thu, Mar 03, 2005 at 03:21:26AM +0100, Henrik Theiling wrote:
> > Only my first conlang Fukhian had three core arguments, and I think
> > that was more a similarity I copied from langs I know than thinking
> > about it. In the next two major projects Tyl Sjok and Qthyn|gai, I
> > also started with three core cases, but instead of keeping them, I
> > dropped them in order to keep the number of grammar rules and ordering
> > constraints low.
> What kind of ordering constraints were you considering?
In Tyl Sjok, I wanted to have an isolating grammar that works by order
of the constituents. I decided that the 'verb' should separate
subject and objects and I chose SVO order. If I had had a third
argument, the problem would have been that two noun phrases would have
been next to each other and I feared that this would be hard to
understand, especially since 'noun noun' is a valid phrase and the
language is pro-drop, so I decided to have serial verb constructions
Eventually, however, I re-analysed the grammar and found that there
was only one open word class and that ordering was by control alone.
I reanalised the parse tree
(agent verb patient)
(agent (verb patient))
thus the agent has control over the verb and patient and and the verb
has control over the patient. To get to only one word class, this
(controller (controller controlled))
Strange system, actually. There is only one grammar rule in Tyl Sjok
now (disregarding the particles that can optionally make the structure
unambiguous) and most of the documentation is my understanding and
interpretation of that rule. It reminded me of Ancient Chinese: the
grammar rules seem vague and most of the time when you try to describe
them, it's actually more like a very long interpretation instead of a
concise description. :-)
> > neither globally, nor for each verb. I personally find the Lojban
> > argument system rather unsatisfactory.
> I've not looked at Lojban in detail, but from what I understand, it
> quite resembles many programming languages in the sense that functions
> (verbs) have a fixed number of arguments that are expected to be
> passed in a fixed order.
Yes, the basic order is
predicate arg_1 arg_2 ... arg_n.
You can drop trailing args without notification, but when dropping
arg_n when you want to mention arg_(n+i), i>=1, you need a special
> A truly verbless lang should not need such circumlocutions.
I'm also not satisfied with these.
> One idea that just occurred to me is to use directional affixes on
> nouns (vaguely similar to Ebisédian), e.g.:
> I kick you --> my_foot you-TOWARDS
> You speak to me --> your_words me-TOWARDS
> The dog runs away --> dog here-FROM
> I leave the house --> I house-FROM
> I look at him --> my_eyes him-TOWARDS
Ha!:-) That looks like Fukhian! :-) By inspiration from Finnish, I had
three spatial cases (for 'at', 'from' and 'to') and from Russian I
stole that the copula could be dropped. You arrive at *exactly* the
same structure above (including word order!), and at least for the
motion verbs, the translation is very similar. Fukhian does have
verbs for 'to kick', though, (and uses them) so most sentences would
have a much more literal sense:
My food goes/comes to you.
My words go/come to your.
The dog goes away from here.
I go away from the house. ('out of' would be: 'away from the
interior of the house')
My eyes to/come to him.
> Hmm, this is starting to look vaguely similar to your system. :-)
Adpositions and verbs are very similar anyway, yes, especially in SVC
languages like Chinese (yong = 'use' or 'with', gei = 'give' or 'to',
dao = 'arrive' or 'towards', etc.) That was one inspiration. :-)
BTW, I just love structures like 'hen3 you3 yong4' = '(it's) very
useful', lit. 'very has use', since this shows how vague word class
distinctions can be. :-)
> ... Every action is rationalized in terms of this model.
So that did work for all verbs?
> Other verbs can be similarly rationalized. So far, I haven't come
> across any verb that doesn't fit into the model in some way.
Really? Hmm, how do you translate:
'I cook water.'
What's the origin? Is this transfer of energy? :-)
'I am tired.'
This is a state, so what concept moves?
> > Maybe it felt like Afrikaans, which has these great negation
> > complements which I copied into Da Mätz se Basa. :-))) Only for
> > each verb, not for negation.
> Does Afrikaans have something similar to the Tatari Faran complements?
> How do the negation complements work?
Quite simple: at the end of a negative clause, you have a final
repeated 'nie' = 'not'.
Dit werk so.
This works so
'This works this way.'
Dit werk nie so nie.
This works not so 'not'
'This doesn't work this way.'
Recently, we discussed some examples here when I was trying to figure
out how *exactly* the structure was.
> > BTW, the evidence markers serves a second purpose in this lang: it
> > marks the start of a sub-clause, otherwise some maybe bad ambiguity
> > could arise. But with evidence as start marker and relative
> > particle at the end, it's properly bracketed.
> Wait, so the evidence markers always begin a sub-clause? So where is
> the matching relative particle for "JIT" in your example sentence
> "John JIT LU Mary MAT KHAN NI" ?
They mark the beginning of *any* clause, so also of the top-level
clause. In an embedded clause, this can be used to determine which is
the first noun.
From your other posting:
> On Thu, Mar 03, 2005 at 03:57:09AM +0100, Henrik Theiling wrote:
> > avoid this reference to roles. A role is simply defined by its verb.
> > There is no generic agent, there is someone who asks. There is no
> > generic patient, but only someone who is hit, etc.
> I like this. But how would you translate something like "what are you
> doing today"? Or, "who did what?" Or, "what did he do to her?" Since
> there would not be a generic agent.
Hmm, good point! I'll have to be careful about generic verbs.
I did not want to say that the concept of a patient is *semantically*
lacking in the language. Only it's lacking in the hard-wired guts of
the grammar (e.g. no specific case-marking for the 'agent' role). And
it should be lacking at every level of the grammar, so that's why I
want to avoid regular derivation processes.
But just like in English, you could identify the entity in a sentence
that is the affected, the causer, the means, the location, etc., and
you could analytically talk about that, thus there will probably be
verbs that specify generic roles. So you could ask for these roles.
And as in English, there will be words for asking for verbs. :-)
The generalisation that has to take place is the same as in answering
'I will *eat*.' to '*What* will you *do*?' in English ('do' is the
generic action, 'eat' is the specific action.)
A generic 'do' will not have two arguments, that's correct, so there
is no direct equivalent of 'to do'. But unary 'to do' will exist,
e.g. expressing a generic 'undertaking', etc. And some other verb
would be 'to be affected by', 'to be an event', 'to be an action' etc.
Let's see the examples:
'what are you doing today'
-> 'Do' probably means 'to undertake' here, so the basic structure
would be something like:
'which INTERROG be-event you undertake this-day happen?'
lit. 'Which event will you be undertaking happening today?'
'who did what'
-> 'which INTERROG PAST-happen?'
-> 'which INTERROG PAST-undertake which be-action?'