Detection and visualization of similarities between social media profiles

Social Media is a generic name given to the digital platforms where users share information. The use of social media has become very popular in recent years. Users share personal information or their ideas about daily events on different social media platforms. The widespread use of social media has led to an interest to extract information from social media profiles and detect similarities among users. If similarity is detected between users, this information can be used to determine the tendencies, interests of users, the things that he/she cares about. Also such an information would make advertisements reach to their target audience. In addition, this information can be used in cases where there is a need to verify the real owner of a social media account. In this study, we have developed techniques to determine the similarities among the users using Natural Language Processing (NLP) techniques. Additionally attempts have been made to identify similarities between users by users' access hours to the platform and users' friends list. However, more effective results were obtained with NLP techniques. Topic Modeling and Named Entity Recognition techniques were used to extract features from the posts of users. Similarities between users were derived by feeding these features to Word Mover's Distance algorithm. Effective results have been obtained using the techniques developed with NLP techniques. It has been shown that communities and clusters in a user set can be detected with the help of these techniques. It has also been shown that social media profiles belonging to a person can be detected by identifying similarities among profiles in different social media platforms. In addition, a web application is developed to make it easier to experiment with the metrics developed in this work.