Mining the Web for the patterns in the Real World

Ever since I first discovered it, I loved the idea of the Moodgrapher and I wish it continued to be developed (not all the functions work any more). What it does is plot, in various ways, changes in "moods" as reported by users of LiveJournal.

For instance, you can see the spike of "ecstatic" on the evening of November 4th, when Obama's victory became official:

i-7c3eb681e1029c7d472ce7b8f0ede113-ecstatic-old.ploticus.moodgraph.jpg

Or you can track cyclic trends - here is "awake":

i-481730c419b6e0a0cad8f359f1a25edc-awake.graph.jpg

I wish the tools could be refined even better, for instance narrowing down the time to just an hour or two or broadening it to months or years.

It would be also nice to narrow it by geography - imagine looking at various happy and unhappy moods displayed on LJ over the course of two hours after the Obama victory announcement narrowed down to country, state, county, precinct or even neighborhood, then comparing this with the official votes? We could figure out, for instance, if conservatives or liberals tend to use LJ more, or to use mood sign on LJ more, or if there are places in the country in which the electoral results are more emotional, perhaps correlated to the narrowness of results or to presence of advertising.

Of course, one can also use Google for similar kinds of tracking. For instance, one can use it to track seasonal events:

i-49bf24e8d224512f48c5084cc614a978-hummingbirds vs. orioles.jpg

Again, it would be nice to be able to set the precise time limits to just an hour or to many years, as well as to restrict them to geographic locations - especially for the appearance of migratory birds and such cases.

Anyone know any more tricks like these, stuff that can be used to track natural or social phenomena by tracking how the Internet responds to them?

Tags

More like this

Looking on google trends I have noticed that a lot of obscure topic searches peak around the time students are writing their term papers.

There was an article recently in EOS (AGU newsletter) on locating and timing the occurrence of earthquakes by surges in visitation to earthquake-related web sites. The surge is prompt enough to permit some localization of the quake's epicenter faster than waiting for the exchange and analysis of seismic data.