الفهرس 
English abstractOnly 14 pages are availabe for public view
 |  | from | 98 |  |  |
| | | from | 98 | | |
Abstract
Sentiment analysis is one of the most important issues that take place now a days. It depends on classifying comments that contain opinions from social network such as (blogs, facebook, discussion groups,..). It is usually used in specifying the percentage of approval or refutation based on analysis of user’s comment. Sentiment analysis is important to users and companies, Although Arabic language is one of the most rich language and becomes is the first language for more than 24 country, studies in sentiment analysis is not enough to show the importance of this issue and its affect in improving the company’s performance because Arabic language has some problems. First, every word in Arabic language has more than a synonym. Such as (ظفر , نصر , فوز, انتصار اوابغض, حقد, كره) . Second, People combine between different perspectives in the same sentence. Such as (هذا المطعم طعامه لذيذ وخدمته جيدة ولكن سعره غالي). Third, Aspects don’t show self strength that affects the final decision of the opinion of being positive or negative.e. g.(هذا المطعم طعامه لذيذ وخدمته جيدة ولكن سعره غالي). This thesis aims at solving the previous problems. As information is increased and rapidly updates, there is a gap between what is taught in the educational institutions and the needs of labor market. To fill this gap, this thesis depends on the analysis of the opinions of student and alumni in a particular domain to enhance the education process. The proposed system depends on a hybrid machine learning algorithms, e.g Support Vector Machine (SVM), Naive Bayes (NB) and Hidden Markove Model (HMM) for Arabic language sentiment analysis. Problems that show that every word in the Arabic language has more than a synonym for that used wordnet used to resolve this problem, people combine different perspectives in the same sentence so aspects used, the aspects of self-force that affect the final decision of the opinion does not appear as positive or negative, so weights used. The results showed that the framework outperforms other systems that use one side of the sentiment analysis were obtained results of up to 95% through the use of the framework, which we have done.