Document Details

Document Type : Thesis 
Document Title :
A Hybrid Personalized Filtering Model for Unwanted Content in Online Social Network Services
نموذج تصفية مشخصنة مختلط للمحتويات غيرالمرغوبة في وسائل التواصل الإجتماعي
 
Subject : Faculty of Computing and Information Technology 
Document Language : Arabic 
Abstract : The high volume of user-generated content caused by the popular use of Online Social Network (OSN) services exposes users to different kinds of content that can be harmful or unwanted. Solutions to protect user privacy from such unwanted content cannot be generalized due to different perceptions of what is considered as unwanted for each individual and due to the differences in cultural values among societies, which are normally shared among users from the same society. Current solutions are either generalized the definition of unwanted content or suffer from the lack of adaptability in the filtering process. Driven by the limitations and challenges in current privacy protection and filtering methods, this work aims at providing a Hybrid Personalized Filtering Model that integrates behavioral, sentimental, and culture-based contextual analysis to detect unwanted content for individuals in OSN services. The hybrid model is proposed to benefit from machine learning, rule-based, and dictionary-based techniques to achieve its goal. In this work, three experimental studies were conducted. The first user study acted as a pilot study that emphasized the demand to solve the problem of unwanted content. The second study involved 394 participants and its aim was to define the classes of unwanted content and to investigate the relationship between the users’ choices of unwanted content and demographic features through correlation and dependency analysis. The final experimental study used Twitter API to extract and process the timelines of a sample of real users to evaluate the proposed model with real data. To validate the final results of the proposed Hybrid Personalized Filtering Model, the outcomes derived from the proposed model were compared with user explicit and semi-explicit feedback using statistical tests. Results revealed an accurate agreement between them. Results also showed that commenting behavior is an effective positive indicator in detecting unwanted content for an individual. Moreover, results showed that age is the most related demographic factor to the users’ choice of unwanted content. The proposed model consists of three classification models: A Sentiment Analysis-Based Classification Model, and Topic-Based Classification Model that uses a supervised machine learning approach, and A Culture-Based Classification Model that uses a dictionary-based approach using the Lucene framework. A variety of Machine Learning classifiers were adopted to achieve the best performance results. Finally, the proposed model can be used in developing personalized solutions to the problem of unwanted content, to increase the Quality of Experience (QoE) for users of OSN services. 
Supervisor : Prof. Arwa Yousef Al-Aama 
Thesis Type : Doctorate Thesis 
Publishing Year : 1442 AH
2021 AD
 
Added Date : Wednesday, September 1, 2021 

Researchers

Researcher Name (Arabic)Researcher Name (English)Researcher TypeDr GradeEmail
مشاعل محمد السلميAlsulami, Mashael MohammedResearcherDoctorate 

Files

File NameTypeDescription
 47172.pdf pdf 

Back To Researches Page