Semantic Analysis of Social Networks Using Hybrid System: Dictionary-Based and Fuzzy Logic رسالة ماجستير

Abstract

In recent years, social networks become an information goldmine provides analyzes and inferences rich environment which can be exploited for the development of knowledge in various fields. The semantic analysis of social media can be classified into three main approaches which are content-based semantic analysis, user-based semantic analysis, and network-based semantic analysis. The first approach is concerned with the content of the posts and mainly on the textual context. The second approach is concerned with the social network users, using the user-based analysis is employing the categorizing of the users according to their patterns usage and the personal trend to have a user-based semantic analysis. The third approach is concerned with user network data such as friends, followers, followees, likes, and shares. In this research, our focus is on the first approach. Several algorithms were used to reach the maximum possible accuracy in the semantic analysis of social networks; the most accurate results were obtained by using the dictionary based and the fuzzy logic algorithms. In this thesis, we worked to obtain better results by creating a hybrid system that fuses the dictionary based and the fuzzy logic to obtain better results rather than using each one of them independently. As a conclusion of the results, we end with a prototype that calculates the polarities of the collected sentences and classify them into seven categories which are Very Positive, Positive, Good, Neutral, Not Good, Negative, and Very Negative in continuous learning manner, the prototype is learning from the previously collected data and changes its previous classifications, which was proven in the results mathematically.