Home

TweetProbe: A Real-Time Microblog Stream Visualization Framework

TweetProbe: A Real-Time Microblog Stream Visualization Framework

Byungkyu (Jay) Kang, George Legrady and Tobias Höllerer

As the importance of social media in our daily life increases, most of people using it witness its significant impact on numerous practices such as business marketing, information science and social sciences. For instance, user-generated information from a few major microblogs are used in order to find consumer patterns on certain type of products. Furthermore, social scientists analyze and predict voting tendencies towards the candidates in a national election. There have been countless research projects conducted on social media datasets in the fields of information science, journalism and so on. In this project, we would like to experiment different visualization techniques with real-time data stream from the major microblog service: Twitter. The project is named as 'TweetProbe' since this visualization framework is designed to present patterns of metadata, topical distribution (in terms of emerging hashtags) and live activities on the current time-window. Particularly, the short time-window used in this project is the key component since it enables users of this application to detect real-time trends, local events, natural disasters and spikes of social signals at microscopic level in time frame.


Figure: Raindrop message visualization



Figure: Raindrop message visualization on a grid (sentiment analyzer enabled)


Figure: A screen shot of the real-time top 10 hashtag visualizer (The spinning box on the right-bottom side of the screen visualizes total count of the newly updated hashtags since the progame has been launched.)



Figure: System architecture diagram

 

The Tweet Stream Probe visualization framework is designed to sense real-time topic-specific trending information on Twitter. In this visualization framework, we implemented both back-end data processing layer and front-end information visualization layer using Java and Processing languages. The first data processing layer filters out unnecessary information from connected tweet stream, updates trending tweets, extracts underlying metadata and sorts tweets, retweets and hashtags. All of these tasks are performed multiple times in each second. We believe that this system can serve social media analysts well for finding useful information or interesting patterns.

 

More info: TweetProbe webpage

 

Relevant Publication