الفهرس 
IntroductionOnly 14 pages are availabe for public view
 |  | from | 178 |  |  |
| | | from | 178 | | |
Abstract
The problem of this study is how to search about the Arabic words in any object (database ”DB”, document, file and so on). These words may be entered as a form inside the object or a variant spelling, as in some names in Arabic language. To solve this problem, we search and analyze the letters and words in Arabic language, and then we recognize the letters that cause the problem. This problem is solved by using four different systems.
First System
This system contains new rules that depended on the previous study in Arabic language. The proposed system is an intelligent system. This system is a new search system which has the ability to understand the different forms of writing.
Second System
The proposed system is designed depending on statistical and probability methods. This system is based on conditional probability and modifies the Hidden Markov Model (HMM).
Third System
It is based on the neural networks that are used to recognize the letters that caused the different forms of the words that were searched. This means that the different forms are represented as one form for each word. In order to solve this problem by neural network, a system is designed using Elman network. The Elman network is considered as a good algorithm dealing with NLP [6].
Fourth System
The first two systems (search by rules and search by statistical) are chosen to search for any Arabic words in documents. Thus, more rules and grammars are added to the first system to deal with this searching. While the second one does not need any modification. Each system does not depend on the other.