I. Introduction
Online social networks (OSNs) develop into an enormous repository of opinions, information, and communications. For instance, Twitter, a micro-blogging service for people to connect with each other and to publish “tweets” (short text messages) has reached 330 million active users in March 2019, and generates 500 million tweets every day or 6000 social signals per second [1], [2]. The emergence of the digital social universe enables our social listening task: to perceive and learn what is happening and what people think (and do) on OSNs. However, like in the real world, the social universe is vast and diverse with unprecedented opportunities to gather voices, it also challenges our ability to sift through and make sense of the vastly noisy signals [3].