Damir, CavarNader, MorkusAlzimam, Mohammed Abdulaziz2026-06-022026https://hdl.handle.net/20.500.14154/79104Treebanks are foundational resources for both theoretical linguistics and natural language processing (NLP), and the theoretical framework underlying their annotation schemes is among the most important factors affecting their quality. For Arabic, existing treebank frameworks are either adopted from those developed for English or other Indo-European languages, such as the Penn Arabic Treebank (PATB) and the Prague Arabic Dependency Treebank (PADT), or partially draw on traditional Arabic grammar, such as the Columbia Arabic Treebank (CATiB). As a result, they exhibit persistent inconsistency in part of speech (POS) classification and fail to capture all Arabic constructions, thereby reducing their adequacy as a representation of Arabic. This dissertation aims to address this gap by introducing the Synthesis Grammar (SG) model as an alternative theoretical framework for Arabic treebank annotation. The SG model is a theory-specific grammatical model of Written Standard Arabic, grounded in the Relative Linguistic Theory, both developed by Mohammed Al Awraghi. The Relative Linguistic Theory classifies human languages into two typological patterns, configurational and non-configurational, and predicts two distinct grammars; the SG model is explicitly designed for the latter, integrating morphology, syntax, semantics, and pragmatics within a single coherent model. The study is executed in four stages: (1) a critical review of the theoretical frameworks underlying PATB, PADT, CATiB, and other Arabic treebanks; (2) a focused exposition of the SG model and its endogenous eight-category POS classification; (3) the design of a corresponding annotation scheme, operationalized in the CoNLL-ASGT format and distinguishing direct relations (extracted from the SG model) from indirect relations (required by the dependency formalism); and (4) the implementation of a prototype treebank, the Arabic Synthesis Grammar Treebank (ASGT), in the INCEpTION annotation tool. The resulting prototype treebank demonstrates that the ASGT annotation scheme can preserve the SG model's layered architecture — compositional structure, combinational relations, parsing structure (cases), denotational relations, and functional structure (thematic roles) — while conforming to the conventions of the dependency formalism. It addresses problems in earlier schemes, provides a comprehensive representation of Arabic, and offers a standardized POS system that Arabic has thus far lacked. The dissertation contributes to linguistics, corpus linguistics, and Arabic NLP on three levels: theoretically, by introducing the first application of the SG model and the Relative Linguistic Theory in computational and corpus contexts; methodologically, by designing the ASGT annotation scheme and CoNLL-ASGT format; and practically, by implementing a prototype treebank intended as the foundation for a large-scale resource.213enArabic treebankArabic Synthesis Grammar Treebank (ASGT)Synthesis Grammar (SG) modelRelative Linguistic TheoryAnnotation schemePart-of-speech (POS) classificationNatural language processing (NLP)Written Standard ArabicCorpus linguisticsNon-configurational languageDependency formalismMorphosyntaxand Syntactic annotation.Thesis