News for hackers. Stuff that matters.
Re:Stocks Plunge, Quickly Recover, on Fake Tweet by pnathan on Tuesday April 23, 2013 @02:33PM (#5597038)
I am writing a tool to scrape social media and do sentiment analysis on that to provide a signal to make stock recommendations. I plan to feed it into an actual trading system at some point.

So the tweet would be integrated into the system, as would any reporting of it. :-/

Re:Stocks Plunge, Quickly Recover, on Fake Tweet by chollida1 on Tuesday April 23, 2013 @02:58PM (#5597207)
How would having twitter integrated into your trading system help you in this instance?

My suspicion is that by the time other people start re-tweeting this the and you can get any sort of twitter sentiment determination the market will have corrected.

Are you taking the other side of this and hypothesizing that you can determine twitter sentiment before the market reacts? If so please mail me, I write in house trading systems for a living and I'd love to chat.

Re:Stocks Plunge, Quickly Recover, on Fake Tweet by pnathan on Wednesday April 24, 2013 @03:21AM (#5600289)
> How would having twitter integrated into your trading system help you in this instance?

(1) In this instance, I'm pretty sure I don't know how to solve the problem naively. It went onto a theoretically legit account. I would have to cross-correlate with a couple other sources of information. At any rate, it points out to me that I need to encode a 'reliability' measure of the information.

(2) Not really. I don't have the capital to operate at low-latency levels. My hypothesis is the obvious one: markets are emotional and I should be able to construct a predictor for certain industries based on emotions emitted on accounts & news snippets relating to those industries.

Re:Stocks Plunge, Quickly Recover, on Fake Tweet by Iterated on Tuesday April 23, 2013 @02:48PM (#5597148)
I'm a finance guy who has been reading about NLP and sentiment analysis. Any introductory books on this subject that you'd recommend for a non-linguist interested in NLP and sentiment analysis? I'm well versed in python/R.
Re:Stocks Plunge, Quickly Recover, on Fake Tweet by pnathan on Wednesday April 24, 2013 @03:29AM (#5600299)
Sorry, I don't have any help, I'm just a regular joe looking at this stuff and taking particularly hacky ways to get the job done on the side (when I don't have other things going on. :-) ). The approach I've been taking initially is the bag-of-words approach. Looking forward I anticipate using some sort of ngram-based approach which clusters sentiment around phrases. These are stupid approaches, but I want to see how far I can push simple and stupid. ;)
Re:Stocks Plunge, Quickly Recover, on Fake Tweet by niggler on Tuesday April 23, 2013 @02:56PM (#5597198)
Warning: There was a fund that tried to do this (https://en.wikipedia.org/wiki/Derwent_Capital_Markets) and didn't exactly succeed
Re:Stocks Plunge, Quickly Recover, on Fake Tweet by RaSoJo on Tuesday April 23, 2013 @02:38PM (#5597072)
Neat. Are you planning to expose any documentation on your tool? Would love to learn how you are approaching it.