On 28 May 2013 13:54, selpa'i <
seladwa@gmx.de> wrote:
>
> la .tijlan. cu cusku di'e
>> lo mlatu ratcu
>> a rat which is related to a cat (e.g. caught or eaten by a cat)
>>
>> la solvor betsemes
>> Betsemes who is related to Solvor (e.g. born to or raised by Solvor)
>
> This brings up a point that has to be laid out clearly. Is a string of cmevla to be considered a single name, or is each cmevla to be seen seperately? Currently, {la .solvor.betsemes.} is just a single name, without any real structure. The Merge must decide whether to keep CMEVLA strings atomic as a whole (Option 1) or to split them apart just like normal tanru (Option 2).
Then multiple adjoining cmevla strings would become a tanru. In fact, if the Merge means that a single brivla or cmevla can equally be a selbri, it wouldn't make sense to keep a group of only the latter type from forming a tanru.
> Option 2 has the effect that some multi-cmevla would undergo a slight change in meaning, e.g. imagine {la .klaus.peter.} whose name consists of two equal parts, neither of which can be considered to modify the other. Or imagine someone whose name contains a glottal stop. The effect that Option 2 would have on such names *might* be negligible, I'm not sure. {la .klaus.peter.} would be a tanru name, so it means a "klaus" type of "peter".
>
> In any case, we could manipulate such names freely. We can {.klaus. bo .peter.} or {.peter. co .klaus.} because once split, the above problem doesn't exist anymore (it's effectively a case of {BRIVLA bo/co BRIVLA}), though it still makes a difference for the meaning of the unsplit cmevla string. It becomes especially apparent once you add a third component that is supposed to modify the entire cmevla string:
>
> lo xekri .ford.taurus.
>
> Option 1 has it that {melbi} modifies {.ford.taurus.}
> Option 2 would interpret it as {(melbi .ford.) .taurus.}
>
> It can't be both, so a decision has to be made.
From my perspective, if koha's family name is "klaus", that's a string not specific to koha but shared by koha & certain family members. It's a generic name, and quite compatible with the semantics of seltau.
If a specific name consists of multiple strings, like "Jean-Claude", and if we didn't want to concatenate them into one string, we could make a tertau that's a pair of parallel selbri joined by a connective: {.jan. JOI .klod.}. Likewise, if a generic/family name has multiple parts, such as "Solvor Camgusmis" in the earlier example, we could again parallelize them -- {.solvor. JOI .camgusmis} -- rather than splitting them into the hierarchy of "first surname" and "second surname".
Or, one could invoke CLL 5:15:
"The tanru may refer to things which are correctly specified
by both tanru components."
>> Or perhaps we could radically change the default tanru order to
>> head-initial, which would be consistent with the right-branching NOI,
>> GOI, ME, NU, etc:
>
>
> Please no. The entire language is generally left-branching, apart from {bo}. Let's not revert something so fundamental.
>
> In what way are ME or NU right-branching? To me, the term "branching" or "grouping" in this context means the grouping of identical (or similar) items in succession, for instance:
>
> broda brode brodi -> (broda brode) brodi
> X .e Y .a Z -> (X .e Y) .a Z
>
> Some constructs are head-initial, but those have two
different components interacting instead, e.g. {KOhA NOI}. I think we
need to dinstinguish between these different situations of branching.
Branching is the orientation of modifiers rather than the grouping of identical/similar items in succession. A construct branches into a particular direction if that's where the modification expands into, i.e. where further information regarding the head item can be found. The construct "preposition + noun" or "tag + sumti" is right-branching, because "noun/sumti" adds information to "preposition/tag" in a dependent manner. And these aren't identical/similar items in succession. Accordingly, "noun + post-position" is left-branching. So is {broda brode brodi}: extra information is piled up toward the left. Below is a useful article on this:
http://www.uio.no/studier/emner/hf/ikos/EXFAC03-AAS/h05/larestoff/linguistics/Chapter%204.%28H05%29.pdf
From the perspective of branching,
ME + sumti
NU + sentence
NOI + sentence
...
are analogous to
tag + sumti
or rather to
selbri + sumti.
ME/NU/NOI/... share the same relative position as selbri, as heads, followed by modifiers: modification expands rightward. Like in English "something is an instance of a cat" (ME), "the cat which ate the rat" (NOI), "that the cat ate the rat" (NU)... Linguists consider these right-branching. And, just like nouns can be a head in relation to a relative clause in natlangs, KOhA (or sumti in general) can be followed by NOI as a modifier, which is again right-branching.
On the other hand, in {broda brode brodi}, modifiers are added leftward: it's left-branching. While some languages are more consistently left- or right-branching, some are less so, which includes English and Lojban. The article explains that "very good" is an exception to the usual English word order of head+modifier because "very", an adverb, is not freely expandable like "which" or "as". However, Lojban seltau is quite expandable - it accepts sumti or tag through {be}:
broda [ be ko'a pe ko'e poi brodo be'o ] brode [ be do'e bei fi'o brodu ko'i be'o ] brodi
This is a left-branching construct containing two right-branching constructs.
> I don't think it's a good idea to get rid of symmetry. The whole point of The Merge is to *increase* symmetry, is it not?
Assuming you are talking about tanru:
Symmetric tanru can be made with JOI/JA. If the seltau-tertau relationship was symmetric by default, we would have to use {ke} etc. for asymmetry. The question is which scheme is more efficient. Are there more symmetric tanru than asymmetric tanru?
Also:
Suppose someone is named "Panz Lanz". "Panz" is the given name, "Lanz" the family name. If we lojbanized and treated each as symmetric parts of the name -- i.e. both {panz} and {lanz} refer specifically to the individual offspring --, we might still want a generic cmevla for different members of the Lanz family. We could reasonably use {lanz} for that. Then what reason would there be to keep {lanz} & {panz} symmetric rather than letting {lanz} modify {panz} as per the usual handling of tanru?
Having family names as independent cmevla rather than an inalienable part of a specific name, can be productive, as in the "Koch brothers", {kok bruna}, "Koch-type-of brothers". "David Koch" could then be {kok deivid}, "Koch-type-of David". I think it would make sense if {kok} in the two tanru referred generically to the same family rather than brothers on one hand and an individual on the other hand, at least to the extent that it *is* a family name. In the "Mario brothers" and the "Doobie Brothers", "Mario" and "Doobie" aren't a family name, but they can still be translated as a seltau modifying {bruna}.
mu'o