Abstract:
Natural language processing (NLP) deals with computational linguistic modeling of large
coverage of vocabulary of human languages. Among the living languages, Semitic languages
can construct numerous lexemes by demonstrating rich morphology, which is an
important branch of NLP. Derivation and inflection, for example, are two such morphological
operations by which Semitic languages can generate numerous inflected or derived
lexemes respectively. Declension is one kind of inflection to construct one form from another.
Among Semitic languages, Arabic is very rich in grammatical declension of nouns
and verbs. In classical Arabic, noun lexemes are declined by nine distinct ways and verb
lexemes are declined by four ways. Modeling the morphological effect of such a rich declension
system is a challenging problem and is essential for intelligent and automated
processing of Arabic language. But this declension phenomenon of Arabic nouns and
verbs has not been captured yet by computational modeling.
In this thesis, we analyze the declension system of Arabic nouns and verbs and design
lexical type hierarchy by which the declension type of any noun or verb lexeme will be
determined from the lexical type. We develop an algorithm to determine declension type
of a noun lexeme. We also show construction rules to capture the morphological and
syntactic effect of declension types dynamically. We also analyze Nominal definiteness
and present construction rules to generate definite lexemes from indefinite. We use Head-
Driven Phase Structure Grammar (HPSG) which provides a versatile, multidimensional,
constraint-based architecture for supporting morphological, syntactic and semantic features
of a language. We show implementation of lexical type hierarchy and construction rules in TRALE, an implementation platform which was developed specially for the grammars
of HPSG. We believe our work effectively extends the capabilities of existing HPSG
framework for supporting declension of Nominals and verbs in Arabic.
