I wrote, with my group, code to use the Bing API, Alchemy API and Quandl API to pull out sentiment analysis and relevance of each article on a company, and then use their stock score to write a retrieval function to score each article based on how well it's predictions came out to be. We used a customized Maxwell Distribution curve and a log penalty. After crawling, we assigned score to all the sources and refined the search result.
