Print

Print


2009/1/11 Sai Emrys <[log in to unmask]>:
> On Sat, Jan 10, 2009 at 11:04 PM, Alex Fink <[log in to unmask]> wrote:
>> For that matter, anyone have any good ideas as to how to make such a thing
>> run smooth?  One reason such a project might fail is if it became too
>> arduous to make a single simple change to the lang and keep everything
>> consistent.  For instance think of sample sentences and texts: you may not
>> be inclined to go through and tweak every one of these which your change
>> ought to effect, and thus they'd gradually fall into disrepair.  What to do
>> about that?
>
> How hard would it be to create a program such that one describes
> sentences meta/interlinearly, and they understand a grammatical
> description, so that a change to the grammar automatically is
> reflected in all generated vocabulary/sentences?

Difficult.

Taking a typical interlinear form and a lexicon, you can produce a
translation without much difficulty. But you're talking about
transformations on that interlinear form. If you managed to do it, I'd
say you could probably publish the results, but I'm not sure how many
people would find it useful.

For example, let's say you eliminate complex gerund expressions in
some situations. These expressions would have to be replaced with
relative clauses. It would take a programmer to describe that one
transformation. And it's not necessarily obvious how to do that
transformation.

There are simple transformations that you can build in -- changing
part of an inflection to a particle, for instance. That would be
useful.

On the other hand, what is the minimal useful thing that could be
implemented? Automatically linking uses of grammatical features to the
relevant grammar descriptions, and vice versa. Maybe automatically
invalidating all usages when the original grammar changes, and
allowing people to mark the examples valid later.

> - Sai
>