Message-Id: <m0sQlZ5-0000YjC@xiron.pc.helsinki.fi>
Date:         Tue, 27 Jun 1995 21:04:03 EDT
Reply-To:     jorge@PHYAST.PITT.EDU
Sender:       Lojban list <LOJBAN@CUVMB.BITNET>
From:         jorge@PHYAST.PITT.EDU
Subject:      Re: TECH: AFTERTHOUGHT SCOPE
To:           Veijo Vilva <veion@XIRON.PC.HELSINKI.FI>
Content-Length: 4532
Lines: 107

> 1. SCOPE-SENSITIVITY
> SCOPE-SENSITIVE terms comprise all sumti whose innermost descriptor
> is (implicitly or explicitly) {lo}.
> SCOPE-INSENSITIVE terms comprise all sumti whose innermost descriptor
> is {le}.

I'm not sure this is right. {ci le reno gerku} is just as scope-sensitive
as {ci gerku}.

I think the way you use them, the scope-insensitive are all sumti whose
outermost quantifier is {ro}. But this is an arbitrary choice, you
could have said that the scope-insensitive are all sumti whose
outermost quantifier is {su'o}, and all your rules would apply
equally, i.e. in {lo nanmu cu batci lo gerku} scope is not an
issue, and for {lo nanmu} with {ci gerku} there are two possibilities.

> At this stage I'm ignoring {la}, {lai}, {lahi}, {loi}, {lohi}, {lei}
> and {lehi}. I'm also ignoring {na} and coordinators.

Yes, we should start with the simple cases, but those should not
present difficulties since they are mostly singular terms which don't
care about scopes. ({loi} is the only exception.)

> Relative scope between two terms matters if at least one of them is
> scope-sensitive. (An exception to this is <{suo lo},{suo lo}>
> and <{ro lo},{ro lo}> pairs: the relative scope of these makes no
> difference to the meaning. I shall ignore this exception.)

Also, if one of the quantifiers is {su'o} or {ro}, then there are
only two possibilities, not three as in the general case. A more
general classification would be:

<{ro},{ro}> or <{su'o},{su'o}> gives only one possibility.
<{ro},non-{ro}> or <{su'o},non-{su'o}> gives only two possibilities.
All other pairs give three possibilities.

But I don't think it is necessary to have rules that depend
on this scope-sensitivity. The cases with {ro} and {su'o} can simply
be left as degenerate cases of the general one, where it just so happens
that some of the three possibilities coincide in meaning. {ro nanmu}
subordinate to {ro gerku} happens to give the same meaning as {ro gerku}
subordinate to {ro nanmu} and also to {ro gerku} coordinate with
{ro nanmu}. No need to have special rules for that.

> 3 KINDS OF AFTERTHOUGHT SCOPE
> The 5 kinds of scope (defined above) are:
>    * X and Y have coordinate scope
>      [X and Y are scope-sensitive]
>    * X has scope over Y
>      [Y is scope-sensitive]
>    * Y has scope over X
>      [X is scope-sensitive]
>    * X doesn't have scope over Y
>      [X is scope-insensitive, Y is scope-sensitive]
>    * Y doesn't have scope over X
>      [X is scope-sensitive, Y is scope-insensitive]
> For the parsimoniously minded, the last two aren't strictly necessary.

Then let's parsimoniously ignore them. Forget about scope sensitivity,
and we are left with:

    * X and Y have coordinate scope
    * X has scope over Y
    * Y has scope over X

> 4 SUGGESTIONS FOR IMPLEMENTATION
> 4.1 Scope relative to another sumti specified by an anaphor
> Add 5 new cmavo in GOI, one cmavo for each of the 5 kinds of scope.
> {SUMTI-X GOI SUMTI-Y} means "X is in the specified scope relationship
> with Y".

This would only work in the prenex, I suppose? Or do you mean to use
VOhAs?

> 4.2 Scope relative to the first/last term in a specified prenex
> Define the following anaphoric possibilities:
>   X is the { first | last } term in the prenex of Y
> where Y, working recursively, is either
>   the { outermost | localmost } bridi
> or
>   the bridi one level of embedding toward the middle of the bridi
>   hierarchy away from Y
> Thus we need the following cmavo:
>   * the outermost bridi - GOhA

Isn't that {no'a}? I never know exactly which bridi this is supposed to be.

>   * the localmost bridi - GOhA [this exists, as {nei}]
>   * the bridi one level of embedding toward the middle of the bridi
>     hierarchy away from ... - LAhE (1)
>   * the first term in the prenex of - LAhE (2)
>   * the last term in the prenex of - LAhE (2)
> [i.e. in total, 3 new cmavo in LAhE, and 1 new cmavo in GOhA]
> These can then be used with GOI as in 4.1.
> The structure would be: SUMTI-1 GOI LAhE (2) (LAhE (1))* GOhA

I really don't understand this part. LAhEs act on sumti. Is that
what you want? Could you write an example?

I think that the coordinate case can be taken care of by adding
a {ro lo} in front of the second sumti. One of the subordinate cases
is taken care of by the standard order, so the only case we are left
with would be solved with something like pc's leaper. Wouldn't that
be enough? This is only considering two sumti, but you didn't look
into the more-than-two sumti case either.

Jorge