A PHRASE STRUCTURE GRAMMAR OF THE ARABIC LANGUAGE

AYMAN ELNAGGAR i

ABSTRACT

A lot of work has been done different from ENGLISH and hence,
in the field of natural language requires a different treatment.
processing (NLP) for ARABIC. Few
researchers have tried hard to apply 2.ARABIC AND COMPUTATIONAL
on ARABIC the methods of LINGUISTICS
computational linguistics, as for As the need increases to a
example, Definite Clause Grammar model of A~%BIC, ARABIC linguists try
(DCG) [I] and Augmented Transition to re-organize the granmlatic system
Networks (ATN) [2]. Because there is in a way that may help the
not a modern linguistic model for researchers in the field of
ARABIC grammar within the frame of computational linguistics. Linguists
computational linguistics, the is traditionally concerned with
results achieved are not comparable formal, general, structural models of
with that achieved on ENGLISH. natural language. Linguists,
therefore, have tended to build
In this paper, we represent the formal models which allow to capture,
phrase structures covering ARABIC. as much as possible, the
They are completely different from regularities of language and make the
ENGLISH. most appropriate linguistic
generalizations. Little or no
i. INTRODUCTION attention has been paid to
Since ARABIC is a rich and characterize the language itself,
highly inflected language [3], there ignoring the m e c h a n i s m that produce
are grannnatical categories in it or decipher it.
ARABIC which do not exist in other
foreign languages such as in ENGLISH So, the researchers spent most
[4]. For example, there is an ARABIC of the time in re-building the models
category 2 called 'I/AFOOL MOTLAK' given by the linguists. Their aim was
which has no equivalent in ENGLISH. to make these models acceptable in
Moreover, there are syntactic computational field (synthesis,
categories in ARABIC which could generation and translation).
themselves be a whole sentence or
even a psepositional phrase An 2.i THE KEY
example of such a category is the What is new in this work is
adverb . In addition, there are other that we classified all the noun
differences in the order of the phrase modifiers into two categories.
constituents within the sentence. For Category one, noun phrase post-
example, in ARABIC, adjectives follow modifiers, in which the ordered
the noun which they modify. sequence of these modifiers is
significant and it is the key to
In fact, ARABIC has its own construct these noun phrase
structure, which is completely structures These modifiers are noun
complement (NC), adjective (ADJ),
apposition (APP) and correlation
i 21 KASER EL-EINI St. CAIRO, EGYPT. (COR). Category two which consists
mainly from VP modifiers like adverb.
2 When there is no equivalent
translation available for any The above categorization has
category in ENGLISH, the given us the ability to focus on each
transliterated form will be shown category and finally the NP
between single quotes. structures were obtained .

342 1
 NP3, NP4 anti NP5 p u b l i s h e d here.
3. THE NOUN PHRASE STRUCTURES
We have Eight structures ACKNOWLEDGMENT
covering all the categories of noun I w o u l d like to thank Dr. T.
phrases. FIVE of these structures are A M B E R (~.) for her valuable discussion
giw~n here. and support Also~, I w o u l d like to
The notation used in thank Dr. H~ MAHGOUB ~) for his
describing these structures is :- comments reviewing the paper and who,
first, implemented most of the
( ) = The category inside these structures given here°
brackets is optional.
*( ) = The category inside these CONCLUS ION
brackets could be repeated more than We have tried here to present
once. the paper in a frame work showing
[ ] = The category inside these that the study of computational
brackets exists once at least° modeling of ARABIC faces many
*[ ] = The category inside these problems. It is evident that:, the
brackets could be repeated but it ARABIC possesses a certain symmetry
should exists once at least° as regards its structure, which leads
itself easily to computation. This
Representation is in work is the basis of any further
Chomsky's Normal Form and should be computational processing of ARABIC,
interpreted from right to left leading to semantic analysis and
eventually machine translation o
The structures given here
consist mainly of NOUN (N) as the REFERENCES
basic unit and its m a x i m u m allowable [i] 'TWO ARABIC SYNTACTIC
post-modifier8 whichcould follow it ANALYZERS', HISHAM EL-SHISHINY &
These post-modifiers have a AYMAN ELNAGGAR, PROCEEDINGS OF THE
specific sequence w h i c h is the key to SECOND CONFERENCE ON ARABIC
construct these structures and COMPUTATIONAL LINGUISTICS, KUWAIT ,
according to the number of these NOV., 1989
post-modifiers, the structures vary
The NP structures as well as some of [2] 'AN EXPERT SYSTEM FOR
the ARABIC categories they UNDERSTANDING ARABIC SENTENCES',
constitute, are given in Fig.(1) M A H E R S. AHMED,
M.Sc. THESIS, CAIRO UNIVERSITY, 1988.
Example of a simple Arabic
verbal sentence is given in Fig.(2) [3] ' AL-NAHO A L - M O S A F F A ',
In this example, the object of the Dr. M O H A M M E D EID, CAIRO U N I V E R S I T Y
sentence has all the post- modifiers
given in the structure NPI and It is [4] 'NATURAL LANGUAGE PARSING AND
clear that some of these modifiers LINGUISTIC THEORIES', U. REYLE & C°
have their own embedded structures ROHRER, D. REIDEL PUBLISIIING COMPANY,
shown in the embedded rectangles 1988

4. MODIFICATION [5] ' A FINITE STATE A U T O M A T A OF THE

'MAFOOL MOTLAK' is a special ARABIC GRAMMAR', AYMAN ELNAGGAR,
modifier in Arabic which has a PROCEEDINGS OF THE 'IEEE
different forms depending on its INTERNATIONAL WORKSHOP ON TOOLS FOR
post-modifiers. We have first [5] AI', GEORGE MASON UNIVERSITY, OCT.,
considered its structure as: 1989.

*(ADJ) (NC) NOUN (~) Professor of computational

linguistics, A L - A L S O N FACULTY, AIM-
But later on, we have found out that SHAMS UNIVERSITY.
both NC and ADJ can not exist
simultaneously and only one of them (2) Project Leader, IBM Cairo
should exist alone. So, we have split Scientific Center°
this structure to three structures

2 343
 CATEGORY NP STRUCTURES

OBJECT
NP1 = (COR) (APP) (ADJ) (NC) N
ADJECTIVE
NP2 = (NC) N
'MAFOOL MO TLAK'
( CASE 1 ) NP3 = N
"k
'MAFOOL MO TLAK'
( CASE 2 ) NP4 = [ADJ] N
CORROLA TION
NP5 = [NC] N

Fig.(1) The NP s t r u c t u r e s with some of the A R A B I C categories they constitute .

ITS PROPER ENGLISH EQUIVALENT IS

THE STUDENT MET THE NEW HEAD OF CAIRO UNIVERSITY Dr. M. SALAMA HIMSELF

F"

"~ ~ , .¢ .~
I ~.~Jtta l

SELF I Dr. M. SALAMA I NEW CAIRO HEAD THE MET

STUDENT

N N L_ N NC

COR APP ADJ NC N N

OBJECT SUBJECT VERB

Fig.(2) Example of a s i m p l e ARABIC verbal sentence

344
3