Frederik Fouvry's HPSG for Dutch

Some time ago, I was working on an implementation for a Dutch HPSG
Grammar.  I had plans to make a description of it, but I never got
round to it, until now (it is still not as I want to have it, but it
is presentable).

It is a short description only, because the grammar can be improved in
numerous ways (wrt to what it can handle).

The grammar was based on the HPSG grammar delivered with ALE.  I had
at first the idea to cover a lot of phenomena, but quite soon I got
rid of that idea ...  I then decided to concentrate on `basic' nominal
groups, and verbal groups without modification.  This last decision of
course made it impossible to test anything on real life input: except
for texts about linguistics, you hardly find any sentences without any
modifications.

After all I have difficulties to be enthusiastic about dazzling long
distance analyses etc. if simple modification is nearly impossible to
treat.

The problems came with the word order difficulties (one of the reasons
to exclude sentence and verb modification) in Dutch. There was
argument inheritance (that brought me the most difficulties
materialised in infinite loops), and the different places some
complements could take (like the participle or prepositional objects).
In order to be able better to cope with the word order, I split up
comps in lcomps and rcomps (I noticed soon after that this is also
what Bouma -- Van Noord used in "A lexicalist approach of the Dutch
verbal complex", but in a categorial context.)

I had to add some rules, because of the word order problems: I removed
schemata 1 and 2, all being replaced by variants of schema 3.  For
head final constructions, I used the solution that is presented in the
manual, viz. to instantiate the cats> list with a definite clause
making the list ending.  This made the grammar less efficient (and
especially so with empty categories), but the results were acceptable.

Verb second is solved in the usual way: with a gap, but the gaps are
not generated by lexical rules.  I changed to this effect the valence
principle and the nonlocal feature principle.  The rules work then as
follows: if there is a verb, then it looks what complements it got.
It is allowed that one is not present at the moment of application of
the rule, and this one will come out of the valence principle and go
into the nonlocal feature principle where the value is unioned with
the inherited slash set.  This handling could be enhanced to insert an
substituting "er" if the subject is extracted, but I don't know if
this is the precise behaviour that is desired to explain the
phenomenon.  (This was about sentences like 
 "Er staat een man op de straat." 
 "There is a man in the street.")

This allowed for a general treatment of all "replacements", like V2,
wh-questions and relative clauses.

The argument inheritance caused major problems, because of the
uninstantiated subcat lists.  Still though, I have spurious ambiguities
in the case of certain verbs that in subordinate clauses allow two
constructions which collapse in the main clause (due to linearisation
effects):

	Hij probeert een huis te bouwen.
	... dat hij een huis probeert te bouwen.
	... dat hij probeert een huis te bouwen.

I did not try to solve this.

Unfortunately the grammar is not working anymore now due to some
improvements I tried to make.  

The description is also on the Web: the URL is
http://clwww.essex.ac.uk/~fouvry/thesis-aoti-english.html 

I hope this is of some use.

Best,

-- 
Frederik Fouvry				  	

PhD student
Department of Language and Linguistics	  
University of Essex			  E-mail fouvry@essex.ac.uk
Wivenhoe Park				  URL http://clwww.essex.ac.uk/~fouvry/
Colchester				  
ESSEX CO4 3SQ				  Tel. +44 1206 87 20 91
United Kingdom				  Fax  +44 1206 87 20 85