Category Specific Post Popularity Prediction

Social media have become dominant in everyday life during the last few years where users share their thoughts and experiences about their enjoyable events in posts. Most of these posts are related to different categories related to: activities, such as dancing, landscapes, such as beach, people, such as selfie, and animals such as pets. While some of these posts become popular and get more attention, others are completely ignored. In order to address the desire of users to make popular posts, several researches have studied post popularity prediction. Existing works focus on predicting the popularity without considering the category type of the post. In this paper we propose category specific post popularity prediction using visual and textual content for action, scene, people and animal categories. In this way we aim to answer the question What makes a post belonging to specific action, scene, people or animal category popular?. To answer to this question we perform several experiments on a collection of 65K posts crawled from Instagram related to four categories: action, scene, people, and animal.