|
Document Type |
: |
Thesis |
|
Document Title |
: |
A TOPIC BASED APPROACH FOR SENTIMENT ANALYSIS USING DEEP LEARNING تحليل المشاعر بالاعتماد على موضوع النص باستخدام التعلم العميق |
|
Subject |
: |
Faculty of Computing and Information Technology |
|
Document Language |
: |
Arabic |
|
Abstract |
: |
In the last few years, Twitter becomes the most popular platform for individuals to share their viewpoints and experiences towards different services and products. Therefore, it attracts a lot of researchers to use it as a body for opinion mining and sentiment analysis. Most of the previous research studies in this area have been using the traditional machine learning-based approach and the lexicon-based approach to classify the emotional states of the tweets in English language. There is limited research work has been done to assort the opinion orientations of tweets in other languages such as Arabic. In addition, recently, deep learning approach has achieved remarkable results over the traditional machine learning algorithms in analyzing a massive amount of data as the case with social networks data. In this thesis, we seek to determine if deep learning approach can be adopted to enhance the sentiment analysis performance for Arabic tweets. Therefore, we evaluated the accuracy performance of existing approaches for sentiment analysis, including machine learning, and deep learning on Arabic tweets to recognize the opinion of the Saudi Telecommunication Companies customers. Logistic Regression (LR), Support Vector Machine (SVM), Random Forest (RF), and Convolutional Neural Network (CNN) have been utilized to detect the sentiment orientation for a dataset consist of 1098 tweets. Results suggest that deep learning technique with Word Embedding method was promising in terms of accuracy (F1=0.81). Moreover, in the aspect detection, results showed that using CNN algorithm for the dataset contained of 1277 tweets achieved the best accuracy reached to 75% when applied with Part of Speech (POS) features. In addition, in this research work, we added a new feature which is extracting the geographical location based on the tweet content. The proposed location model reported the following scores 0.6 and 0.89 in term of precision for point of interest and city respectively. |
|
Supervisor |
: |
Dr. Amal Abdullah AlMansour |
|
Thesis Type |
: |
Master Thesis |
|
Publishing Year |
: |
1441 AH
2020 AD |
|
Added Date |
: |
Monday, June 22, 2020 |
|
Researchers
| نوره فهد الشمري | Alshammari, Norah Fahad | Researcher | Master | |
|