الفهرس 
Thesis MotivationOnly 14 pages are availabe for public view
 |  | from | 157 |  |  |
| | | from | 157 | | |
Abstract
This thesis shows different comparative studies of opinion mining using machine learning algorithms on various domains such as social media (Twitter and Facebook), smart entertainment (IMDB), product and devices reviews (Amazon), and some different Arabic datasets. Also, the thesis presents the process of constructing an Arabic tweets dataset that can be used for opinion mining. The steps of constructing and preprocessing this dataset are also discussed. The preprocessing includes removing non-Arabic letters, tokenizing, removing stop words, removing repeated characters, removing URLs, removing user mentions, removing hash-tags, removing retweets, removing diacritics, handling emoticons, normalizing Arabic analogous letters, labeling the tweets into polarity classes and handling skewness.
The thesis proposed two methodologies for Arabic and English text opinion mining. The common stages for both methodologies include feature extraction and machine learning based classification. The feature extraction algorithms that are used in the studies are term frequency, term frequency inverse document frequency and n-gram. The machine learning algorithms that are used in the studies are Naive Bayes, Multinomial Naive Bayes, Bernoulli Naive Bayes, Support Vector Machine, Logistic Regression, Stochastic Gradient Descent, Adaptive boosting, Maximum Entropy, Passive Aggressive and Ridge Regression.