![]() | Only 14 pages are availabe for public view |
Abstract In a Biological Brain no clear distinction exists between memory and reasoning. The above idea has been a motivation of the present work in which we aim at solving a problem , that usually needs the generation of a set of rules that simulates the knowledge of an expert in the domain-by using the associative retrieval capabilities of the connectionist approach. Specifically, the objective of this study is to implement an Arabic morphological analyzer using associative search techniques via hamming neural network memories. This implementation requires an exhaustive list of all possible Arabic generalized standard forms (GS Forms), and requires the existence of a lexicon that stores all possible roots Hamming memories are used to store the Available GS forms and the lexicon roots. The input to the system is a given Arabic word the output is GS Form that best matches the input word, And the corresponding lexicon root. This study contains six chapters, and three appendices. Their contents are as follows: Chapter I: presents introductory material in natural language understanding (NLU) and some associated processing problems. Chapter II: presents an introduction to computing with neural network Chapter III: present a numerical representation method for Arabic words Chapter IV : is the core of the present work , It introduces a neural based approach for Arabic morphological analysis. Chapter V : presents the implementation of the developed S/W package to simulate the proposed neural-based system. Chapter VI : concludes the present work and discusses some possible future extensions. |