Quantifying News Sentiment To Predict Stock Performance
posted on: 2009-11-19 16:17:35 by Soo-Young
tags: technology

---

Today I happened upon a site that really caught my interest called StockMood.com. As the name suggests, StockMood tries to determine the mood of a stock based on news stories using, what they call, "natural language processing models". How cool would it be to be able to use a website to get a read on all of the news and information out there about a particular stock. With the accessibility of news on the internet, I'm surprised something like this hasn't come along sooner. I signed up for beta testing, was granted access, and this is what I found. Full disclosure: this is not a paid post and I have no affiliation to StockMood.com.



The first quote I tried was GOOG and was surprised when I saw the price displayed as 282.75 (GOOG is currently trading for 572.99). I looked more closely at the pricing information and noticed that the date, which read: January 20, 2009. So obviously their pricing information isn't current, but that's not a big deal because they're still in beta.

The first thing I noticed on the quote page were the stock charts, which included price, mood, sentiment moving average, and positive/negative articles. The charts are really great, especially the price chart because it includes an overlay of upgrades and downgrades of particular stocks (the two main sentiments given by StockMood are: High Optimism, and High Pessimism). You can see how the price changes with each sentiment rating.

I looked through a few quotes and found that the sentiment indicators weren't especially foretelling, although the price charts seem to be limited to a 3 month period, and StockMood only seems to be pulling stock news from one source. I'm looking forward to seeing more chart customization options as well as more news sources added to each stock feed. With more information, I'm sure StockMood will be able to pull out more useful information for investors.

The challenge StockMood is taking on is enormous. I can't even begin to imagine how to program a script to accurately parse through the news to reach a meaningful conclusion about a stock. I'm definitely going to keep my eye on this website and see where it goes in the future.