Re: bridi, tanru and lujvo

[Logical Language Group <lojbab@ACCESS.DIGEX.NET>]

la .eduard. TCR,lin. cusku di'e

> I am a beginner at lojban, but not at math and computer science. I hope
> this is useful:

[excellent explanation of relational algebra deleted]

> Is there a reasonable, provably unambiguous way to represent arbitrary
> relational join-projection operations in lojban, as bridi?

There is, indeed.  Predicate logic is a superset of relational logic,
which is based on a carefully restricted form that's computationally
easy to implement.

> If not, how could we make one?

Here's how to assert rigorously that Mon Repos is a doghouse for Spot:

        da poi me la spot. de poi me la mon. repos. di zo'u
                da gerku di .ije de gerku da
        (Ex)(x is Spot) (Ey)(y is Mon Repos) (Ez):
                (x is-a-dog-of-breed z) and (y is-a-nest-for x)
        There exists an x (Spot) and a y (Mon Repos) and a z such that
                x is a dog of breed z and y is a nest for x.

> Can this be spoken and understood, in general?

So we hope!

> Can we apply the same idea to tanru?

We could, but it vitiates the purpose of tanru, which is to be able
to speak NON-rigorously.  If, in order to express an idea which is
not yet lexified, we must firmly establish what the exact relationship
is that we are expressing, we will have trouble using the language in
a spontaneous and natural manner.

Similarly, there are other areas of deliberate semantic ambiguity.
Notably, names are speaker-relative, not pinned down to some specific
onomasticon, so we cannot know who "la djan." is without context.

> Can we apply it to the generation of lujvo, where we might want to make
> more than one lujvo from a particular pair of gismu?

Remember that lujvo are not limited to gismu >pairs<; they can be
arbitrarily long.  As we say in Lojbanistan, the price of infinite
precision is infinite verbosity.

> Can we define lujvo explicitly on the fly as a sort of extensible
> pro-bridi, as we make definitions in mathematics or carry out relational
> operations against a DBMS, or do we have to go through a formal procedure
> and haggle over the details each time? I personally hold with Humpty Dumpty
> on this: "When I use a lujvo, it means exactly what I wish it to mean,
> neither more nor less," because I wish to be able to say explicitly what I
> wish it to mean, and I foresee a need to make up lots of them.

It is a tanru, rather than a lujvo, which is Humpty-Dumpty-esque.  The
point of lujvo is to freeze useful interpretations about which there is,
or ought to be, consensus.

--
John Cowan              sharing account <lojbab@access.digex.net> for now
                e'osai ko sarji la lojban.