This is fantastic work. I have been thinking about this for a long time. We have even built a few demos.
I think this is a great start to show the power of WL. I was thinking of adding other data sources such as Google Trends because search trends are likely related to sentiment.
You could cross reference the articles with search trends to come up with a possibly more rich data set.
Obviously, this could additionally be cross-referenced with Twitter trends and other text or numerical data.
I was also thinking you could parse the articles a little deeper with TextSentences and other extractors, then run them through sentiment as well.
Sentiment could be partially analyzed via Twillers geo metadata. Knowing what geo locations are swinging the markets could potentially yield some fruit as well.
This post is inspiring though because this is such a hot and exciting market. Great work!