Retrieving Relevant and Interesting Tweets During Live Television Broadcasts

The use of social TV applications to enhance the experience of live event broadcasts has become an increasingly common practice. An event profile, defined as a set of keywords relevant to an event, can help to track messages related to these events on social networks. We propose an event profiler that retrieves relevant and interesting tweets in a continuous stream of event-related tweets as they are posted. In our application, these tweets are to be displayed in real time along with the live broadcast of an event. To test our application we have executed a user study. Feedback is collected during a live broadcast by giving the participant the option to like or dislike a tweet, and by judging a selection of tweets on relevancy and interestingness in a post-experiment questionnaire. From the experimental results we conclude that our event profiler is capable to find relevant keywords which leads to retrieval of a higher number of relevant and, on average, more interesting tweets than using only a manually selected keyword.