ABSTRACT
INTRODUCTION Microblog services have emerged as an essential platform for people managing interpersonal relationships with friends, posting updates about daily activities, publishing and exploring messages of their personal interest. Twitter [1], one of the largest worldwide microblog service platforms, has 1 billion registered users, and over 200 million active users send 400 million microblogs per day [2]. Due to the open data policy and abundant application programming interfaces (APIs), researchers and companies can crawl data from these platforms and conduct analysis for their own purpose. Actually, microblog services have become a manifold Big Data [3] source for analyzing people’s relationships, daily thoughts, comments, and interactions on particular concerns, like TV programs. As reported by Nielsen [4], a third of active Twitter users tweeted about TV-related contents during June 2012, which refers to an increase of 27% from the beginning of that year. Mining social media contents associated with TV programs, resulting in a novel paradigm of social TV analytics [5], can extract many insights to fulll multiple purposes, such as oering targeted advertisements, interactive program composing, user marketing, and so on. Inspired by the commercial success of Bluene and Trendrr [6], it has been widely known that social TV analytics can benet the whole TV ecosystem, from TV content producers, TV channel operators, advertisement agencies, to audiences. erefore, building a Big Data platform for social TV analytics attracts more and more attention from both academia and industry.
