To make it work, we take a sample from Twitter every 30 seconds and analyze them in 50-result batches for associations and term matches. They accumulate for 5 minutes and then we flush sample aggregates to the database. So the database has samples from when it started through present in 5-minute granularity. As new terms trend, they begin to populate on the X-axis. The system back end will be implemented using a Java-based stack and PostgreSQL RDBMS. The presentation will be implemented using Flash targeted to Player 9, standards-compliant XHTML/CSS targeted to modern browser versions with significant market share (Safari 2+, Firefox 2+, IE 6+).
Auswertung der Nachrichten bei Twitter, um einen Überblick über die U.S. Wahl zu bekommen.