Twitter Social Networking Graph using Hypernym based Semantic Similarity Detection

With the increasing number of people taking to social media, it has become important that the people get introduced into their networks of people they know and their interests. The role of recommendation systems is to create more connections for people which would create interest for people to develop their profiles. Social interaction graphs show how users on social media handle Twitter, follow each other on social media. The proposed model aims to create a social interaction graph based on the similarity of the content of tweets. Hypernym based feature engineering is performed to improve the semantic analysis of tweets. The dataset contains tweets from users and up to 15 levels of hypernyms are added. The similarity between tweets is evaluated using k-Nearest Neighbours model and potential users to follow are recommended. 39.73% of recommendations are actual connections thereby showing the effectiveness of the system. The social networking graph of different users created by the proposed system is visualized in 2D space.