[matrix] matrix - general: inserting chunks into the type hierarchy

[Emily M. Bender]

Dear Gertrud,

With some delay, I am getting around to your question no 2.
>From the description below, it sounds to me that Northern Sotho
is very similar to many of the Romance languages, in that 
verbal arguments can be expressed by affixes which are
orthographically treated as independent words.  If that's
right, my suggestion would be to attach them to the verbs
via a preprocessing step, and then treat them as morphology.
(Independent noun phrases, on the other hand, are syntax,
and thus recursive, as you note.)

Hallmarks of affixes include fixed order and lexical
idiosyncrasy of various kinds.  For the argument that French
so-called "clitics" are really affixes, see Miller and
Sag 1997 "French Clitic Movement without Clitics or Movement"
in NLLT.

As for how to implement this in a Matrix-based grammar, I
will send you a grammar of Italian that one of my students
created.  It doesn't do any of the clitic climbing, but it
does show how to do the basic clitic attachment.

Emily

On Mon, May 21, 2007 at 12:16:31PM +0200, Gertrud Faasz wrote:
> this is the second of a number of questions on the type hierarchy of the
> grammar matrix, each we will post separately, for they touch different
> issues.
> 
> question no. 2 concerns chunks
> 
> Northern Sotho verbs are written disjunctively i.e. morphemes carrying
> morphosyntactic information on verbstems are usually preceding the stem
> as stand-alone units. Concerning the object noun phrase(s) usually
> succeeding the stem, one can see the following phenomenon:
> 
> the "ordinary case":
> 
>  <NP monna> <VP <V o reka/> <NP dipuku>> - (the) man he buys books
> 
> The object noun phrase can be replaced by an object concord which will
> precede the verb stem:
> 
>  <NP monna> <VP <V o a di reka>> - (the) man he MORPH_pres them buys
> 
> (The present tense morpheme only occurs if nothing follows the verb)
> 
> Moreover, some object concords are merged to the verb stem.
> 
>  <NP mosadi <VP <V o bona>> <NP monna> (the) woman she sees (the) man
> 
>  <NP mosadi <VP <V o a mpona>> (the) woman she MORPH_pres him-sees.
> 
> One example of two objects (there's another nice phenomenon: a VP
> syntactically constitutes a complete sentence in Northern Sotho - the
> subject NP is only contributing adjunct semantic content and is often
> left out):
>     
>  <VP <V o mphile> mpho> he/she me-gave (a) present.
> 
> 
> This is only one example of such an deletion/inversion operation
> concerning the object. Another is the reflexive. To solve the problem,
> we defined a set of nine "verbal chunks with/without an Object" called
> VCO. That is that before any other analysis on the verb is performed
> (like, e.g. negation), we put the verbstem together will all required
> objects into this chunk. In the following examples, we all call them VCO
> for the sake of simplicity, actually the names are different for each of
> the structures.
> 
> <NP monna> <VP <V o <VCO reka/> <NP dipuku>>> - (the) man he buys books
> <NP monna> <VP <V o a <VCO di reka>>> -(the) man he MORPH_pres them buys
> <NP mosadi <VP <V o <VCO bona>> <NP monna>>(the)woman she sees (the) man
> <NP mosadi <VP <o a mpona>> (the) woman she MORPH_pres him-sees.
> <VP <V <VCO o mphile> mpho>> he me-gave (a) present.
> 
> We'd like to define these structures not as phrases for they are not
> recursive. 
> 
> Now the questions:
> Is it better to define "chunk" as a subtype of phrase or shall we rather
> add it parallel in the hierarchy (We would prefer the subtype) -  how
> can we define that chunks are not recursive?
> 
> If there is any example matrix grammar available describing such a
> phenomenon we'd be happy to be allowed to take a look.
> 
> Thanks a lot!
> Gertrud