A Semantic Graph Based Micromodel to Predict Message Propagation for Twitter Users

摘要

Twitter is the most popular social platform for broadcasting opinions, but the reach of tweets is often nondeterministic. Recent years has witnessed numerous agencies misusing the platform for entrapment techniques, psychological manipulation and fake news campaigns compelling Social Media firms to enforce stricter data protection policies with limited access as the norm. This paper presents a micro-prediction model for determining message propagation for a user, especially for the non-influential majority. Our framework uses Ego network and Named Entity Recognition in predicting message propagation. The work focuses on determining the possible users who would interact and their immediate reach. This is achieved by using Twitter API in a limited manner. We attempt to make a responsive prediction model; simple, stateless and scalable, capable of catering to parallel requests. The simulation predicts with an accuracy of 85% for data constituting 336768 connected users.