1 Comment

Summary:

Bloomberg now has sentiment analysis tools to tell traders about spikes in positive or negative chatter on Twitter.

Twitter goes public on the New York Stock Exchange. Photo: Getty Images
photo: Getty Images

Financial firms once blocked Twitter to prevent employees leaking sensitive information. It worked, but there was a problem: reporters and companies were using Twitter to reveal important news, leaving traders to rely on second-hand accounts.

“If Carl Icahn tweets about Apple, you can’t wait for some journalist to write it up,” said Brian Rooney, who is Bloomberg’s head of product for news. “As social moved into the financial sphere, it became absolutely clear we had to track it.”

In response, Bloomberg introduced a Twitter compartment that sits alongside the flow of other news and financial data that appear on its famous terminals. The feature, which provides a way for employees to read but not send tweets, appeared last spring, shortly after the SEC announced that public companies like Netflix  can use social media platforms to disclose market-moving information.

This month, Bloomberg took  the technology a step further by announcing a sentiment analysis tool that not only alerts traders of a spike an activity, but also shows if the news is likely to be bad or good. Here’s a screenshot of the tool in action after a CNBC reporter tweeted Comcast acquiring Time Warner Cable, 2014’s biggest business deal to date:

Bloomberg chart re CMCSA

The chart shows velocity alerts (in purple) that reflect a burst in social media activity for a certain stock, as well as bursts of positive (in green) or negative Twitter sentiment. According to Rooney, velocity and positive news alerts for Time Warner Cable appeared on the Bloomberg screen within 90 seconds of the initial CNBC tweet.

Twitter’s positive bias

To detect the mood about a stock on Twitter, Bloomberg relies on sentiment analysis tools that the company developed in-house. Rooney says that tasks like machine learning, auto-translation and natural language are all core competencies for an information company like Bloomberg, which has 3,000 engineers. Like other analytics companies, Bloomberg pays GNIP and Datasift for access to Twitter’s firehose of information. (To hear other examples of how big data tools are changing markets and industries, come join us at Structure Data in New York City on March 19-20.)

Bloomberg has discovered that news about companies on Twitter is disproportionately positive, according to Rooney. One reason is that language and expressions about “a deal” are generally neutral or favorable.

“We have found that Twitter is biased to the positive, but that can make negative tweets more important,” he said, adding that Bloomberg may tweak its sentiment algorithms to account for the bias.

Overall, Twitter’s impact on the market also reflects how the age of big data is changing the habits of finance professionals.

“The challenge used to be ‘where do I get information?’ Now, it’s about curating and making the information actionable.”

You’re subscribed! If you like, you can update your settings

  1. Pure sentiment analysis has been around for a while from both Bloomberg and Reuters. Interesting to see Bloomberg pushing it out there a little more now and it will be interesting to watch how they handle elements such as the overly positive bias amongst other nuances of the data.

    Christian@Trendensity

Comments have been disabled for this post