Artificial Intelligence: The Answer to Wall Street’s Data Deluge

Big Data may have met its match with a new wave of artificial intelligence software aimed at cutting through the clutter.

supercomputer-large.jpg

Would artificial intelligence technology have helped JPMorgan CEO Jamie Dimon and his lieutenants catch and grasp the full story behind the firm’s approximately $6 billion trading losses — buried in the trading data — far faster?

Is a narrative presentation that swiftly provides insight about a large body of financial data preferable to investors and financial managers than is the presentation of pure visual data in a chart or spreadsheet format? Is a word-based explanation more quickly understood and thus more quickly acted upon?

While these questions continue to be debated, new technology developed by Narrative Science, a two-year-old firm born out of artificial intelligence research conducted at Northwestern University, is poised to use machine-generated narratives employing AI to help Wall Street manage its exponentially growing deluge of data.

At a recent technology conference in New York targeting the financial services industry, Narrative Science demonstrated its new AI-based software platform, called Quill, that aims to help individuals in the financial sector and other high data-generating industries to better analyze and comprehend the full meaning of the data more quickly.

“Someone would come by and at first think that all we are doing is text aggregation, and then it would hit them: We take existing data, raw financial data — it can be about investments or data they need to analyze or data they need to communicate to clients — and quickly turn it into stories, a narrative or executive summaries,” that appear to be written by humans, explains Kris Hammond, chief technology officer of Narrative Science.

During a question-and-answer period at the conference, Hammond asserted that his AI platform can conduct swift analysis that can also be used to detect data anomalies, highlight financial fraud and rapidly issue a verbal communication about the findings — sort of like have an investigative reporter or data investigator on site.

But he was also quick to qualify the firm’s capabilities: “We work with structured data — numbers and symbols — where the numbers and symbols are unambiguous. We do not work with free text,” he said. “We can’t take a news article and generate a report, but if there’s data where we have unique symbols for a company and a link to its current earnings, sales or market cap info, we can take all that data, do an analysis and generate a story.”

This is the opposite of what a host of other companies now do, which is to take unstructured data, such as information from the Internet, social media, blogs and tweets, and try to aggregate and make sense of the data. Narrative Science flips that model, working exclusively with structured data, the kind of numbers- and symbol-based data financial firms largely revel in, and uses algorithms, machine learning and a natural-language generator to turn that data into fluid content so that it reads as though it has been produced by a human. Rob Passarella, a financial data specialist, describes the technology as “a platform that takes large amounts of structured data and turns it into narratives, so that humans can make sense of the complexities inside.”

Narrative Science’s unique approach to addressing the “Big Data Challenge,” as the dilemma is sometimes referred to, comes from the company’s unusual genesis: Hammond, the founder of Northwestern’s AI laboratory, decided early on to partner with the university’s respected journalism school to test and refine his lab’s capabilities. A collaboration between engineers and journalism students ensued, allowing the two groups to apply advances in artificial intelligence to content and news generation. At first, the technology was used to produce sports articles based on local baseball team statistics. “You would be amazed at how much analytics is required to produce a really good sports story,” notes Hammond. But in time, he and his partners realized that the technology, while relevant to media companies — some of whom are now clients — was also relevant, if not more so, to data-driven businesses like finance.

The Evanston, Illinois–based company was founded by Hammond along with chief scientific advisor Larry Birnbaum, another Northwestern computer science professor, and CEO Stuart Frankel, previously an executive with digital advertising company DoubleClick, a Google subsidiary.

Related