Some things

Motherfucker! It seems that between the time Sycara and Thomas wrote their papers on using text data for financial prediction, apparently the Yahoo finance boards went to shit. The MSFT discussion board is one big never-ending flamewar about Microsoft vs. Linux -- very little discussion of Microsoft's financial status actually occurs. Remarkably, a cursory analysis of the discussion surrounding heavily traded stocks like GOOG, MSFT and so on reveals that actually there's very little information in the posts. It's not intelligent, thoughtful discussion of the kind you'd find at Fool.com or the Wall Street Journal, it's mostly typo-ridden indecipherable garbage that the most sophisticated natural language processing methods are powerless to deal with. Yahoo finance board discussion data is, in other words, so ridiculously noisy as to be useless.

last updated 2 years ago # history