الفهرس 
IntroductionOnly 14 pages are availabe for public view
 |  | from | 201 |  |  |
| | | from | 201 | | |
Abstract
Anaphora is a linguistic phenomenon occurring in all natural languages and it means pointing back or referring back to an entity in a text. Anaphora resolution is to bind each anaphor (Pointing back word or phrase to its antecedent (the entity to which a word or phrase refers) in a text.
Although anaphora is a challenging process in Modern computational linguistics studies, it is very important as a lot of Natural Language Processing (NLP) applications need a success anaphora resolution to achieve their goals in machine translation, text summarization for instance.
The objective of the thesis is, thus, to present a new syntactic model or algorithm for anaphora resolution in Arabic, which can be applied in computerized applications. The proposed computerized system follows a rule based approach to solve Arabic anaphora resolution depending on Arabic LDC Treebank part 3 version 2.This system depends mainly on the theory of constraints and preferences in selecting the correct antecedent of an anaphor.
The proposed syntactic rules are generated specially for Arabic and compatible with its specific structures as will be mentioned in following chapters. The study is an attempt to solve anaphora resolution in Arabic. The proposed system achieves a success rate of 86%.