Advanced Place Stucture Mangling

[Jacob Errington]

P.S. (as in pre scriptum ?) This e-mail turned out longer than I had originally intended, but I think that it's worthwhile letting the mailing list in on this discovery I made just yesterday and expanded today in #lojban. It gets technical at times, but I have provided translations for all the examples, which I have tried to use as much as possible to make things more tangible.

We all know that {be} works on the selbri-level, allowing us to inject sumti directly into the selbri. We also know that the primary use for this is in description sumti, where we can select an x1 based on definite (i.e. non-zo'e) values in other places. (We also probably agree that {lo broda be fa ko'a} is either nonsense or equivent to {lo broda no'u ko'a}, but that's outside the scope of this e-mail.) However, being done on a selbri-level, what's really happening with {be} is that we're creating "new" selbri each time we use it.

Consider {lo klama be la .bastyn.}: we create a new selbri with the definition "x1 goes to Boston ..." in order to select the x1 and get a referent and whatever. (The actual way articles work and possible issues with that is also outside the scope of this e-mail.) What's really interesting, however, isn't description sumti. (In fact, that's pretty boring.) What's really interesting is *what* is in the "..." of my previous definition of {klama be la .bastyn.}. The full definition is "x1 goes to Boston from x2 via x3 by means x4."

Notice how the x3 place of {klama} has become the x2 place of {klama be ko'a}. Indeed, injecting a sumti has the effect of moving all the later places *forward*. What repercussions does this have on top-level bridi?

Consider {dunda}, "x1 gives x2 to x3." Given the above proof of place promotion, {dunda be ko'a} should mean "x1 gives ko'a to x2." Considering that {ko'a broda ko'e} equals by definition {ko'a ko'e broda}, {mi do dunda be ko'a} means "I give you ko'a," mirroring the English structure! By using {be} and the ability to move places into the bridi-head, we can create pseudo-"1 3 2" argument order at the cost of *one* syllable. The actual SE conversion required to achieve the true structure is  {se te se} which is three syllables long and requires forethought. In a way, this selbri-level manipulation can almost be thought of as "afterthought SE conversion."

We can create even wonkier place structures by using {be FA}. Basic {se} can be achieved using {ko'a broda be fa ko'e}, e.g. {lo nu lo pampe'o cu darno cu badri be fa mi} "My lover being far is what saddens me." The real complexity of the transformation comes from moving bridi-tail sumti into the bridi-head: {.i mi do lo barda cu vecnu be lo plise} "I you for a lot sold an apple." (Mirroring the Lojban with English leads to ugliness, but I'll try to keep it up so long as it's understandable.) The argument order in that case is 1-2-4-3, with a corresponding SE conversion of {te ve te}.

From this, I concluded the generality that when all sumti are moved to the bridi head, the effect of {broda be fa xi ny ko'a} is to send the x_n place to the end of the place structure. In a way, this can be thought of as "remote" FA, as it allows us to perform a FA operation on the selbri level, outside the formal place structure.

An example, with corresponding argument order and SE conversion: 

{.i lo tcadu zdani lo nurma zdani cu klama be fa mi} "To the city house from the country house go I." 2-3-1 {se te}

Before continuing, we must recall that {se} is applied before {be}, as {lo se broda be ko'a} has {ko'a} in broda1.

Combining this type of {be} operation with ordinary SE conversions, we can produce extremely cryptically ordered selbri:

{.i lo ni se pluka lo ka pinxe lo ckafi kei kei do te zmadu be mi) "In the amount of enjoying drinking coffee, you are greater than me." 3-1-2 {te se}

What is truly shocking about this is that we see a {te}-conversion, but the selbri becomes reduced into a binary predicate, i.e. a predicate with two argument slots, because of the {be ko'a}.

Everything above is what I had discovered yesterday. What follows is what I discovered today.

It is (currently) ungrammatical to use {be} twice on the same selbri. (Not that I'd necessarily want it to be.) Indeed we have {bei} for that. But, we can think of the following as "forethought bei." 

Using {ke} to create a bracket, we can box one selbri inside another, {ke broda [ke'e]} has the {broda} selbri trapped inside the {ke..ke'e} selbri. The {ke..ke'e} selbri has the same structure as the inner selbri, and this is what we can exploit to avoid using {bei}. Formally, {ke broda ke'e be ko'a} is equivalent to {broda be ko'a}, but if broda already has linkargs, i.e. injected sumti with {be}, we can't move the outer {be} inside. 

Consider {klama be la .bastyn.}, "x1 goes to Boston from x2 ..." We can't attach another {be} to this selbri because it would be ungrammatical, but we can box it inside {ke..ke'e} and make use of the place structure transparency outlined in the previous paragraph and *then* use {be}: {ke klama be la bastyn [be'o] be la montre'al}, "x1 goes to Boston from Montreal via x2 in vehicle x3. In cases where the first sumti does not end in a selbri, the use of the second {be} will cause (a lot) of elision. 

Although "forethought" {bei} might appear utterly useless at first, it can be used to avoid using multiple FA. Suppose we want to specify the x3 then the x2 with injected sumti. Normally, we need to use {broda be fi ko'a bei fe ko'e} because that's the way it works when you only have one selbri, but when we use ke..ke'e as a selbri "box", we get to cheat, by considering that the inner selbri has its own place structure:

{.i mi fi lo karce cu ke klama be fi la .montre'al. [be'o] [ke'e] be la .bastyn.}

First, we consider the inner selbri {klama be fi la .montre'al.} as having the place structure "x1 goes to x2 from Montreal via x3 in x4."

Then, we box that selbri inside the ke..ke'e brackets, and use be again, to fill the x2, i.e. the destination, with {la .bastyn.}. This yields the place structure "x1 goes to Boston from Montreal via x2 in x3."

Finally, when we do formal place filling, we use *{fi}* to specify the vehicle, in this case {lo karce}

"I, in the car, go from Montreal to Boston."

Although using ke..ke'e by itself to create a place structure-transparent "box" is pretty interesting, we can create even what I consider to be the most advanced structure changes by using a combination of all the tools outlined above: SE, {ke..ke'e}-boxing, {be}, and {bei}.

In sum, I have showed some ways that we can achieve more complicated argument order by usually using {be}, moving sumti into the bridi head, and using some FA in weird places, and how we can cheat the system with "forethought bei" by using selbri-boxing.

Any comments, thoughts, violent objections, hate mail, or flame wars are appreciated as usual iu

.i mi'e la tsani mu'o

[la gleki]

banli sai mu'o  u'i

[selpa'i]

la tsani cu cusku di'e

> P.S. (as in pre scriptum ?) This e-mail turned out longer than I had
> originally intended, but I think that it's worthwhile letting the
> mailing list in on this discovery I made just yesterday and expanded

> today in #lojban. [...]

An interesting read, but I'm not sure you can call it a "discovery" if
it's been in jbofi'e all along (or is jbofi'e's output what you based
your discovery on?):

{mi do dunda be ko'a}:

[1(2[dunda1 (give-r(s)) :] mi I, me)2 (3[dunda3 (recipient(s)) :] do
you)3 [is, does] �4{5dunda giv-ing be (6[dunda2 (gift(s)) :] ko'a
he/she/it)6}5�4]1

Clearly, someone must have implemented it this way (knowingly, I
assume). Of course, we cannot be sure that jbofi'e has any authority
about such things. The most useful interpretation should optimally be
the preferred one.

Here is a past thread on this very topic:
https://groups.google.com/forum/#!msg/lojban-beginners/QiZ5pF9q34I/MDaYSI0q0agJ

It doesn't look like there was a consensus of whether "ko'a ko'e ko'i
broda" should mean the same as "ko'a ko'e broda be ko'i". Personally, I
think it should be as jbofi'e parses it (that is, they are not
equivalent), because it gives us more options/flexibility.

mu'o mi'e la selpa'i

[ianek]

I'm sure I've talked about this here before.
It's true that {da de broda be di} is equivalent to {da broda di de}, but it's not true that {broda be di} has different place structure than {broda}. It's just that the second place is already filled, so {de} takes the next unoccupied place. Meanwhile, {da fe de broda be di} is equivalent to {da broda de fe di}, so there are two sumti for the second place (possibly equivalent to {da broda de jo'u di} or something).

At least that's how it is now, I'm not sure whether you're talking about current Lojban or proposing new things.

mu'o mi'e ianek

[Jacob Errington]

On 24 January 2013 16:07, ianek <jan...@gmail.com> wrote:

I'm sure I've talked about this here before.
It's true that {da de broda be di} is equivalent to {da broda di de}, but it's not true that {broda be di} has different place structure than {broda}. It's just that the second place is already filled, so {de} takes the next unoccupied place. Meanwhile, {da fe de broda be di} is equivalent to {da broda de fe di}, so there are two sumti for the second place (possibly equivalent to {da broda de jo'u di} or something).

At least that's how it is now, I'm not sure whether you're talking about current Lojban or proposing new things.

I don't know if you could really call it "the way Lojban is now" because stuff like this is extremely rare. What I'm trying to do here is provide a comprehensive analysis of what happens on the selbri level when linked arguments are used. The conclusion that I reached is that {be} is a selbri operator much like {se} that will indeed *modify* the place structure. {broda} and {se broda} are indeed different selbri, even though one is clearly derived from the other. Similarly, {broda be ko'a} and {broda} are *different* selbri. 

(I have put parentheses around the selbri in places where isolating it aids comprehension.)

With that said, it shouldn't be unusual that the place structures be different. Because using {be} drops a place from the structure, the other places should move forward. {ko'a ko'e (broda be fo'a)} is equivalent to {ko'a (broda be fo'a) ko'e}  because we are allowed to move sumti from the head into the tail freely. The sumti that are linked directly into the sumti are invisible to the formal structure sumti. This becomes more obvious when we isolate the selbri and write this out in {me'au} notation. {me'au} is an experimental cmavo that allows function abstractions (i.e. {ka}-abstractions) to become selbri.

e.g. {mi do (me'au lo ka ce'u cinba ce'u)} -> {mi cinba do}

(Formally, {me'au} is of selma'o ME.)

Do demonstrate the effect of be operating on a selbri level, using me'au helps, such that {broda be ko'a} as a selbri is equivalent to {lo ka [ce'u] broda ko'a [ce'u] ...}

Therefore, {.i fo'a fo'e (broda be ko'a)} -> {.i fo'a fo'e (me'au lo ka ce'u broda ko'a ce'u)}

Using {fe} on the top level therefore has no effect, as it is operating on the formal place filling level and not the selbri-internal level, where {ko'a} resides in the example just above.

{.i fo'a fe fo'e (broda be ko'a)} -> {.i fo'a fe fo'e (me'au lo ka ce'u broda ko'a ce'u)} which is equivalent to {.i fo'a fo'e me'au lo ka ce'u broda ko'a ce'u}. The {fe} indeed has no effect of double-assigning a place, in this case. Another way of looking at it would be to say that using be moves the sumti directly into the selbri, which is outside the scope of top-level FA. Even if at this point it were still debatable, saying that it causes double assignment would be less productive, I think, as double assignment in general is unuseful. (I don't think it's entirely pointless; it's one solution to the problem of using {sei} and splitting quotes: {.i lu coi pendo li'u selsku lo verba fa lu .i do mo li'u})