Message-Id: <199408150836.AA01035@nfs1.digex.net>
Date:         Mon, 15 Aug 1994 11:34:48 +0300
Reply-To: veion@XIRON.PC.HELSINKI.FI
Sender: Lojban list <LOJBAN@CUVMB.BITNET>
From: Veijo Vilva <veion@XIRON.PC.HELSINKI.FI>
Subject:      Nested preposed relative clauses
To: Bob LeChevalier <lojbab@access.digex.net>
Status: RO

I am taking a wider view at the problem of preposed relative
clauses seeing that Jorge isn't very happy with them.

We can divide languages in two classes: to those which use
descriptors (articles) and to those which do not. English,
Spanish, German and Lojban belong to the first class, Finnish,
Japanese and Chinese to the second one. I don't know Spanish
or Chinese so I speak nothing about them - I only make inferences.

Some descriptor languages (English) have very limited possibilities
of using preposed restrictive clauses. German fares better but
succumbs soon with a bad case of center-embedding, as does Lojban.

In some non-descriptor languages (Finnish) the preposed clauses
are a viable/preferred alternative to postposed relative clauses.
In some others (Japanese) they are about the only possibility.

  J:  ((((machi) e iku) otoko) o miru) kodomo
            N       V
            |_______|
                   VP     N
                    |_____|
                         NP        V
                          |________|
                                  VP     N
                                   |_____|
                                         NP

  F:  ((((kaupunkiin) menevan) miehen) nakeva) lapsi
  G:  das ((den (in die Stadt gehenden) Mann) sehende) Kind
  L:  le (poi (le (poi (le tcadu) cu se klama ku'o) nanmu) cu se viska
      ku'o) verba
  E: *((((into the town) going) man) seeing) child

The German version is still almost useable while I find the Lojban one
unpalatable (and Jorge finds it indigestible). The Japanese and Finnish
versions are OK as there is no EXPLICIT center-embedding.

Lojban uses modifier-modificand order in tanru but has presently no
viable way of doing it in cases like the above one even though probably
about half of the world's population is using similar structures -
including at least one of Lojban's root languages.

The preposed relative clauses are a rather late addition to the
Lojban syntax. At the time we were discussing it probably no one
looked into the case of several nested clauses. Even a cursory look
would have indicated a need to handle them properly. It is possible
to do it in Lojban with a minor amendment in the grammar (now I have
grokked it, I think). There is no problem re {poi/noi} as the clauses
seem always to be restrictive (at least in Finnish/Japanese).

The grammar of sumti tails would have to be augmented with the
following rule:

  *| XOI sentence [KUhO sumti_tail_1# [term#] ... [CU#]
     bridi_tail ] ... XUhO sumti_tail_1

This would keep the rest of the grammar unchanged at the cost of
introducing two new selma'o/cmavo. The example sentence would become

  *le xoi le tcadu cu se klama ku'o nanmu cu se viska xu'o verba

The other possibility would be to replace the rule

   | relative_clauses sumti_tail_1

with

  *| preposed_relative_clauses sumti_tail_1

and add

 * preposed_relative_clauses = GOI term /GEhU#/
 *   | NOI sentence [XUhO sumti_tail_1# [term#] ... [CU#]
 *     bridi_tail ] ... /KUhO#/

This would require but a single new selma'o/cmavo {xu'o}. The example
would read

  *le poi le tcadu cu se klama xu'o nanmu cu se viska ku'o verba

The type of NOI would define the nature of the outermost clause,
all the inner ones would be restrictive. Replacing {poi} with {noi}
in the example would give a meaning equivalent to

     le verba noi viska le poi klama le tcadu ku'o nanmu
or   le verba noi viska le nanmu poi klama le tcadu

No language known to me is able to handle {ke'a}s unambiguously
in the case of nested relative clauses - whether preposed or
postposed. In Lojban, one route out of trouble might be to use
assignable KOhA:

  *ko'a goi le poi ke'a klama ko'a le tcadu xu'o nanmu cu
   se viska ku'o verba

Here {ke'a} = {le nanmu} and {ko'a} = {le verba}. In Finnish the
referent of ko'a would have to be inferred from context which
cannot always be done unambiguously. In Lojban, a KOhA can be
assigned at each EXPLICIT level of nesting (even the postposed
relatives will get slightly unpalatable but at least you can
achieve unambiguity if absolutely necessary):

  ko'a goi le verba poi viska ko'e goi le nanmu poi ...

The basic structure of Lojban is such that there is no way to avoid
descriptors. However, the above amendment would remove the innate
center-embedding problem in one of the most frequent cases and
enhance the useability of the language significantly (from the
viewpoint of people with non-descriptor native tongues, e.g. quite
many non-Indo-European ones.)

This amendment would be just a finishing touch to the previous
change which added the preposed relative clauses (which were
deemed necessary to mend the otherwise broken relatives).

  co'o mi'e veion

---------------------------------
.i mi du la'o sy. Veijo Vilva sy.
---------------------------------