Pointless weekend project idea: A service that scans articles for matching regexes to classify the level of bullshit of what you are reading. It could even be a browser extension with the bullshit indicator automatically shown for each page in its icon.
This just requires some NLP to identify obvious hyperbole. In fact, you could probably just watch HN for long enough, establish some training data and have this happen somewhat automatically for you in the browser.
Ran it on this(http://www.naturalnews.com/049944_vegetable_oil_omega-3_pest...) NN article and it's still processing 20 minutes later. I'd like to think it's breaking whatever scale is being applied to it, but it's most likely the connection speed in my office. :P
