Is Shoebox4.0 good for Dakotan?

[Koontz John E]

On Tue, 31 Aug 1999, Constantine Xmelnitski wrote:
> I've been trying to urge the Shoebox to parse Lakhota
> texts properly yet there's one stumbling block for it
> to do well:
> You seemingly cannot arrange affixes into slots,
> preventing, say, pronominal affixes appear after
> reflexive or dative ones.

It's been a while since I looked at the Shoebox parser, but it has always
been rather weak, requiring a lot of human intervention to function.  Note
that the parser, while useful, is a long way from being the main reason
one would recommend Shoebox, it's just there to help automate glossing
somewhat.

> Also we cannot block some ungrammatical affix
> combinations.
> As a result I have either a dozen of ambiguities (with
> all the variants of -wa-, -ya-, -ki-, etc.) or
> morpheme trash.

Siouan morphology tends to produce ambiguities, due to the many
homophonous morphemes.  At least theoretically, position class analysis
isn't a general solution, as it the consituent classes may vary in order
depending on the identity of the morphemes.  Furthermore, some words
consist, essentially, of two (or more) sequences of subwords defined in
terms of similar sets of position classes.  A few classes are shared by
such subword sequences, e.g., inclusives.  Adding some derivational
prefixes can modify (supercede) the positional structure of the underlying
stem. Some of these assertions are less true with Dakotan than, say,
Dhegiha, but they are true of all MV languages.

> Perhaps there's some solution?

I'd just expect to have the system come up with numerous cases in which it
needed to have me tell it which wa or ya the current one was.  The
expectation is that you have to gloss most things to the extent of picking
them out of a list.  If you get some non-root entries for free be
satisfied.

I had the impression that of the several SIL parsers available, the
old-fashioned and somewhat complex AMPLE looked like it might be the most
likely to work well with Siouan morphologies.  However, it doesn't
integrate with Shoebox as far as I know.  I think you'd have to parse
outside Shoebox, and I'm not sure if this external parsing could be
imported usefully.

If anyone has any experience with this, I'd be delighted not hear it.  I'm
rather rusty with Shoebox myself; I've never actually used AMPLE, I've
done very little parsing of Siouan data with the Shoebox tools, none
since, I think, version 2.