Mining the Twitter Stream: Unravel Events, Interactions, and Communities in Real-Time

Under the topic of event detection, this paper reviews and discuss different mining and visualization techniques that could provide good insights about how to detect real life events in the twitter stream: a lightweight event detection using wavelet signal analysis of hash tag occurrences in the twitter public stream, description of detected events using a Latent Dirichlet Allocation topic inference model based on Gibbs Sampling. A visualization of the live twitter feed with diffusion networks constructed using user relations like the 'Retweet', '@reply' and '@user' mentions, and large scale dynamic community detection allow an effective detection and tracking of the communities in the public stream. Paper is a compilation of author's previously published work in the areas of mining and visualizing the twitter stream.