News Arbitrage

So I've heard of high frequency trading before, and had an idea, and was wondering if someone smarter than me can tell me why it doesn't work.

So a little while ago this paper came out from the University of Arizona that managed to data mine Yahoo Finance articles for words that would correlate with stock performance. It seemed to be moderately successful. Full article here:

http://ai.arizona.edu/intranet/papers/Textual%20A…

Then there's this other site here: http://newsmap.jp/ This is basically a neat way of representing news stories with the size of the box the story is in proportional to the number of hits the story has received.

So this is my thought: Combining the two ideas, it doesn't seem so hard to come up with an algorithm that would trade based off of key word data mining articles that weights the importance of each news story with the number of hits received. With a little machine learning and basic statistics you could isolate the cyclicality of the news cycle from trend.

This isn't exactly high frequency trading, but it does seem like a way of setting up a money machine fairly easily. I can't believe that some trading house like Getco or the like doesn't do this unless the idea is pretty bogus.

Any thoughts?