The impressive TweetMap ALPHA [], developed by Harvard University’s Center for Geographic Analysis, is based on a dataset of about 95 million tweets, which can be dynamically queried by time, by location or by keyword.

Tweetmap makes use of so-called special Graphic Processing Units (GPUs) to parallelize the hard-to-compute job such as the querying and rendering of the very large dataset on-the-fly. As such, the online map shows individual tweets as unique dots, but also aggregates them as a heatmap. In addition, the a series of maps can be generated and explored by time, ranging from days to mere seconds.

See also TweetPing: Revealing Worldwide Twitter Activity in Real Time.