[SHOEBOX] answers to David and jeff

At 06:22 21.7.2000 -0400, The Gray Wizard wrote:

>Since I have effectively created a [SHOEBOX] tagged thread here, I have a
>question for other Shoeboxers out there.  amman iar has an auxiliary verb
>form that is semantically empty, serving only to specify the grammatical
>usage (mood [prefix], aspect[infix], tense [suffix]) of the lexical verb and
>as such has no root form, only affixes.  Shoebox will refuse to parse a word
>without a root.  As a work around I have arbitrarily selected the mood affix
>(which is always required and happens to be a prefix) as the root form.  Is
>there a better way to do this?
Does the verb actually not *have* a root, or is it that being semantically
empty it cannot really be glossed?  Assuming it is the latter, it would
seem a good idea to gloss it as 'AUX'; the use of _do_ in English often
approaches this, as in:
     Does        John    read        books?
     AUX -s      John    read -0     book -s
     AUX -3.SG.  PN      V    -INF   S    -PL

As you can see I make use of the null morpheme //0// as a device to hang
tags on.  I imagine you could provide a lexical form which de_fact consists
of only affixes and endings with a null root.  I've never really done this
in Shoebox, but I did once play with the idea of a Romance artauxlang
having the three auxiliaries
     ar  < HABERE    -- forming perfect
     er  < ESSE(RE)  -- forming passive and progressive
     ir  < IRE       -- forming future
whose roots //a e i// thus coincided with the theme vowels of the three
conjugations.  I would probably have made the morphemic analysis along the
lines of //a+ar, e+er, i+ir//, since the verbs in question arguably had a
"meaning" of sorts, but IMOO //0+ar// &c. would have been valid as well.


ho Jeff Sheets egraphe de:

>Certain things that Shoebox doesn't handle gracefully in the alternate
>forms bit is recognizing vowel harmony.  At least I haven't found a good
>way to do that, except to input all alternate forms. Blegh.
In one book about Finnish morphology that I have they list alternating
morphemes like -sta~-stä 'out of' with an underlying form //-stA//, letting
the //A// morphophoneme being changed to a/ä at the
morphophonemic--phonemic level.  Look in the help file for "Preventing
Incorrect Parses" and "Morphophonemics" for hints on how it may be
done.  One way of implementing this practically might be to tag all
back-vowel roots in the lexicon with an underlying form adding a later
deleted dummy affix //-H// which triggers the right affix form.  More
elegant still would be a rules file which progressively change //AOU// into
/aou/ after preceeding /aou/ and /äöy/ otherwise.  Look up the topic
"Reduplication", "Infixes" and "Context Sensitivity—Preventing Incorrect
Parses" in the help file for suggestions how VH may be handled more
elegantly, and look at "Adapt3a", where The notation "C*" is used for zero
or more instances of "C".  (I've tried to look where the * convention is
defined, but couldn't find it, so presumably it is hard-wired into Shoebox.)

It might look like:
         \lx -stA
         \a -sta
         \a -stä
         \u stA
         \ps ninfl
         \ge ELATIVE
         \ge out of

I would use a rule along the lines of the one in Adapt3a:
         Shortening of vowels in verbal nouns (31)
         \ru [C][Vl][C][C*]#
             [C][Vs][C][C*]#
Which may be adapted thus:
       Finnish Vowel Harmony
       \ru [Vb][C*][V]
           [Vb][C*][Vb]
       \co Vb = back vowels

I'm positing a greater number of underlying alternants/vowel morphophonemes
than actually occurs at the surface.  VH ignores /i/ and /e/, for which
reason it is convenient to posit back vowels //ï// and //ë// which merge
with the front //i e// at the surface level.  Incidentally this is an
actual historical sound-change, as shown by the fact that Estonian has a
mid back unrounded vowel {õ} corresponding to Finnish /e/ in words that
used to have back harmony, like Est. _põld_ vs. Fi. _pelto_ 'field', also
some Finnish /i/s do cause t_ > s, and some /i/s don't -- presumably the
latter are //ï//.  I prefer taking front vowels /äöy/ as basic and /aou/ as
secondary, since words containing only the "neutral" vowels /ie/ take front
vowel affixes.
      It occurs to me that such an approach may work also with umlaut, by
positing underlying //ï// and //ë// (historically *i: and *i), causing
umlaut and then merging with //e// and zero respectively on the surface, as
in German //mann+ïr// -> _Männer_ (pl. of _Mann_) and //grab+ët// ->
_gräbt_ (3.sg.pr. of _graben_ 'dig').  (Note that the Shoebox docs talk
about "underlying forms" while I prefer a historical approach "*a became *ä
before *ï/*ë, then *ï was lost and *ë merged with *e, leaving /ä/ as a
phoneme".  The net result is the same, IMHO.)
      Again I have had no occasion to try any of this myself, so I don't
know if it really works.  Actually I don't know how one should make Shoebox
make the right assosciations between members of the [V] and [Vb] sets --
possibly by defining  a set of Harmony Vowels [Vh] as suffix vowels [Vx]
with following w, so that VH turns //AOU// into //Aw Ow Uw// after [Vb],
with later rules cleaning up AOU+w into aou and AOU without following w
into äöy.  This suggests that -w- should be handled as an infix at some point.

kai h'autos Jeff men:

At 17:00 20.7.2000 -0400, Jeff Sheets wrote:
>In a message dated 7/20/00 12:36:31 PM Central Daylight Time, bpj@NETG.SE
>writes:
>
> > Has anyone attempted implementing soundchanges in Shoebox?  So far all my
> >  attempts have implemented headache in me ony!
>
>My current project takes root phonemes and applies a sound changes rule file
>to them, which more often than not drastically changes the pronunciation.  In
>my previous post I gave the example of:
>
>\uf feg - qin - lo - as - yu
>\sf feqrinlez
>\mg admit - future - complete - I - PL
>
>You can see that the combination of g-q becomes qr, the combination of o-a
>becomes e, and the combination of s-yu becomes z.
I've mulled over this.  It seems to me you would actually want a rules file
like "\Shoebox\Samples\Adapt\Adapt2b\EngPhon.rul" for this:

    \ru g -q
        qr
    \ru z -yu
        z
and so on.  In this way you won't need to list all possible morpheme
sequences in the lexicon, I think.



/BP
--
  B.Philip Jonsson <bpX@...> <melroch@...>
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
"Truth, Sir, is a cow which will give [skeptics] no more milk,
and so they are gone to milk the bull."
                                    -- Sam. Johnson (no rel. ;)