Date: Tue, 17 Jan 1995 20:29:16 -0500 (EST)
From: jorge@PHYAST.PITT.EDU
Subject: Re: replies mainly re "ka"
Sender: Lojban list <LOJBAN@CUVMB.BITNET>
To: Veijo Vilva <veion@XIRON.PC.HELSINKI.FI>
Reply-to: jorge@PHYAST.PITT.EDU
Message-id: <01HLZ69BUBXI000TK6@FIPORT.FUNET.FI>
Content-transfer-encoding: 7BIT
MIME-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Content-Transfer-Encoding: 7bit
Content-Length: 6528
Lines: 169

And:
> > >       feature      feature value
> > > (I)   colour       red, blue, green
>
> A feature is a selbri, a function from the possessor to the value.

I agree that it is a function, but it can't be a selbri if we are going
to put it in the x3 of frica. It has to be a sumti.

Let's first agree on what {frica} means. I'm interpreting it like this:

        ko'a ko'e frica *feature*
        Koha and kohe differ in *feature*

Say the feature is colour(). Then:

        ko'a ko'e frica *colour()*
        Koha and kohe differ in colour.

means that colour(koha) is not equal to colour(kohe), i.e. the function
colour() takes different values for the arguments koha and kohe.

This means that x3 is not the actual difference between x1 and x2, but
the characteristic in which they differ. For example I could say that
"li 3 frica li 5 in magnitude", but not that {li 3 frica li 5 li 2}.

(The gismu list seems to accept this last possibility, but let's at
least agree that it corresponds to a different meaning of frica.)

With that assumption, it is clear that we need a function in x3.
Then something like:

        ko'a ko'e frica le se skari

would be wrong, because what is in x3 is not a function but a specific
colour (the one the speaker has in mind).

To get a function, I would use:

        ko'a ko'e frica le ka ke'a skari makau
        Koha and ko'e differ in what colour they are.

The {ke'a} is necessary to show which of the many vacant slots is the
one that corresponds to the argument of the function being evaluated.

The {makau} is necessary to show the output of the function. Without it
they could differ in who perceives them with some colour, or under what
conditions they are perceived to be some colour, or even more generally,
on having the property of being coloured.

On the other hand, neither {ke'a} nor {makau} is absolutely required
if context makes it clear, so that I would be happy to say:

        ko'a ko'e frica le ka skari

and leave {ke'a} and {makau} as obvious if context allows.


> I don't think redness is a feature: "what is the redness of this
> book? Answer: crimson" does not exemplify the normal meaning of
> redness.

I can think of redness either as a binary function (or feature), taking
values "red" and "non-red" (then the redness of a blue object would be
"non-red"), or as a multivalued function with values crimson, vermilion,
and what have you. I would probably understand this last one in the
sentence "A and B differ in redness", i.e. the function redness()
evaluates to something different for A than for B.

I see {le ka ke'a xunre} as the function redness(), where {ke'a}
simbolizes the variable, so that the function is not evaluated.
{le ka ko'a xunre} on the other hand, is the value that the function
redness() takes for the argument koha.

Without further context, {le ka ke'a xunre} is a function that takes
the values "red" and "non-red". {le ka ke'a xunre la'u makau} can take
many values, "amount of redness" however you want to interpret
that, and can be written more concisely as {le ni ke'a xunre}.

> > > "Differ" needs a feature (e.g. size) as x3, not a feature value.
> > Agreed.
> Right. So I don't think "lo ka broda" is adequate as x3 of "differ".

How else can you get an unevaluated function there?


> > If you can do that then there is no problem. In the case of {ka}, just
> > add a rule that if the bridi is all filled, so that there is no room
> > left for {ke'a}, then a {do'e ke'a} is assumed.
>
> That rule is fine. But it doesn't solve a more awkward problem: when
> more than one syntactic tersumti are unfilled, which is to be filled
> by keha, and which by zohe?

I offer two possible rules: 1- The first vacant tersumti is filled with ke'a.
2- The obvious tersumti is filled with ke'a.

These will probably coincide a lot of the time, and usually at x1.
I'm inclined towards rule 2, but it doesn't really matter, either one is
reasonable.


> > ta goi ti
> > vo'u zunle vo'o
>
> I don't see the problem with these two - I mean I don't see why they're
> semantic garbage.

{goi} assigns a value to an assignable  pro-sumti. I don't know how you
interpret {ta goi ti}, but it has to be some generalization of that which is
not obvious to me.

{vo'u zunle vo'o} is also meaningless because {zunle} doesn't have x4
and x5 slots. Of course your rule of taking them as {zo'e} saves us, but
that is cheating a bit.

> > zi'o poi klama le zarci
>
> I see the problem with this, and it could perhaps be fixed by changing
> the syntactic status of ziho. (Personally I'd have preferred 5 cmavo
> in SE, meaning "zap the nth place".) But this problem is not too
> bad, because only the poi clause is uninterpretable; it doesn't affect
> the bridi it occurs in.

Maybe a computer can easily discard the clause, for a human being it would
be much harder to ignore it.

> > li pipaipi
>
> If that ends up uninterpretable then it ought to be fixed.

The members of PA behave semantically quite differently one from another.
To achieve syntactic and semantic balance we'd have to split them into
several selmaho, but I don't really see a need for that.

> Every grammatical Lojban sentence ought to be interpretable by a
> computer: that, to me, is at least implicit in the promise of
> Loglan/Lojban. At the least there should be general rules of
> repair, such as "if a sumti is uninterpretable, replace it by
> {zohe}".

How about a more general rule: if something is uninterpretable, ignore it.
It covers your rule for sumti.

> By the logic of your "le ni keha clani", one could also have
> "la djan frica la meris le duhu keha clani" - D & F differ in
> terms of which of them is tall.

I would use {le ka ke'a clani} for that, which would be the binary
function tallness() with values "tall" and "non-tall". Since it
evaluates differently for each of them, one has to be tall and the
other non-tall.

I really don't like {du'u} there, but at the moment I'm not sure why.

> Ka is "the property of being (an) X", and lihi is "the experience of
> being (an) X", but ni is not "the amount of being (an) X". Rather,
> ni is "the amount by which Y is (an) X". So I agree ka and lihi
> should be together (outside NU), but ni should stay with nu.

That is playing with words, you could equally well say that ka is "the
property by which Y is (an) X", and li'i "the experience of Y of
being (an) X".

What they have in common is that they can work as unevaluated functions,
which is what some tersumti require (like the x3 of frica).

Jorge