Detecting Opinions in a Temporally Evolving Conversation on Twitter

The immense growth of online social networks from simply being a medium of connecting people to assuming a variety of roles has led to a massive increase in their use and popularity. Today, networks like Facebook and Twitter act as news sources, mediums of advertising and facilitators of socio-political revolutions. In such a scenario, it is of vital importance to be able to detect the opinions of social network users in order to study the opinion flow processes that unfold in these networks. For many topics, the focus of the conversation evolves over time based on the occurrence of real-world events, which makes opinion detection challenging. Since it is not practical to label samples from every point in time, a general supervised learning approach is infeasible. In this work we propose a temporal machine-learning model that has its underpinnings in social network research conducted by sociologists over the years, to detect user opinions in evolving conversations. It uses a combination of hashtags and n-grams as features to identify the opinions of Twitter users on a topic, from their publicly available tweets. We use it to detect temporal opinions on Obamacare and U.S. Immigration Reform, for which it is able to identify user opinions with a very high degree of accuracy for a randomly chosen set of users over time.