Twitter Aurora
with The Office for Creative Research
Aurora (originally called Manifold) is a live, immersive experience revealing the breadth of conversations on Twitter. The top 300k user accounts are clustered into interest groups based on the Twitter follower graph, which can be explored in three dimensions.
Trend detection in those clusters provides an alternative to the geography-based trending topics shown on twitter.com. It’s an alternative landscape that ignores political boundaries and instead highlights how people speak within interest-based communities.
Photos
Technical Detail
The data behind Manifold is both live and static. The follower graph of the top 300k Twitter users (by follower count) were run through t-SNE by Twitter to provide a 2D embedding, after which I used DBSCAN to create clusters which were re-arranged into via a custom interface the structure that you see in the final product.
The live data analysis listens for hundreds of tweets per second from these top accounts, then does trend detection on hashtags and phrases found in those tweets, revealing which topics are actively being talked about in particular clusters. Videos about those trends can be exported as well.
The wand technology was developed by Oblong Industries.
Process
Prototypes and studies from the initial development of the project.
