I wanted to analyze a long-running group conversation on Google Hangouts that I'm a member of. It would be see where the message activity was anomalous, and be notified automatically when it's worth checking the conversation.
What it does
It will feed the number of messages per hour (or some other aggregation interval) in the conversation over time to NuPIC. NuPIC will analyze the data stream for anomalous events, and produce a graph of the anomalies in the data over time.
This kind of project could be used to monitor any online conversation between people (like on news sites, communities, etc.), to automatically generate notifications when the conversation is anomalous and thus interesting to check.