This content is from: Portfolio

Mining Data From Trader Jargon

Cloud9 Technologies partnered with Google and Quantiphi to build a machine learning-powered voice transcription service for traders.

For forty years, traders have called each other up to price potential transactions. Now, a fintech firm wants to turn those calls into market data.

Through a partnership with Google and data science firm Quantiphi, New York-based Cloud9 Technologies has developed a transcription service that can translate verbal communications between traders into text data using machine learning. This data, the firm believes, can revolutionize compliance as well as offer new insights into trading activity.

“It’s going to be transformative for the industry,” said German Soto Sanchez, Cloud9’s global head of corporate development.

Although the majority of equities traders have shifted to electronic trading, the bulk of fixed-income transactions are conducted verbally, with traders calling peers at other firms to get price quotes and ultimately agree on a transaction. Roughly 90 percent of fixed-income trades are done this way, according to Cloud9.

But the advanced phone systems used for voice trading have not really changed over the last four decades, Sanchez said. He said Cloud9 was founded in an effort to modernize and improve these systems – an effort which now means recording and mining conversations for data.

From a compliance point of view, turning voice recordings into data makes it possible to “pick out patterns and even emotion,” said Jerry Starr, co-founder and CEO of Cloud9. “Hopefully, catch a trader before he makes a bad trade,” he said.

[II Deep Dive: Financial Regulators Embrace Artificial Intelligence]

Sanchez added that the tool could also improve trader workflow and offer an additional source of market data. Although voice transcription technology has been available in other sectors for a while now, Cloud9 said they are the first to come up with a version that works for trading, given the brief, jargon-filled nature of communications between traders.

“It’s another language, it’s vernacular-heavy, the pace is very rapid,” Sanchez said. “Because of that nobody has really been able to transcribe it.”

By leveraging Google’s machine learning technology, Quantiphi co-founder Asif Hasan said they were able to build a new transcription model from the ground up that can understand traders.

“Voice trading is becoming more pronounced, more tech savvy,” Sanchez said. “To leverage that and provide deeper analytics of trader activity – it’s a very exciting frontier we’re in.”

Related Content