Cloud9 announces super-fast transcription of trader talk
Although traders have been awash in data for years now, very little voice trading data has been susceptible to analysis. But last month, Cloud9 Technologies announced that it had cracked the code.
The new system can transcribe nearly 100% of voice trades accurately, far better than prior attempts, according to Germán Soto Sánchez, global head of corporate development at the New York firm. “Historically, a number of firms have tried to do that, but it’s been very difficult for them,” he said. “They haven’t been able to get the accuracy beyond the low 60s.”
One limiting factor faced by earlier experimenters was the quality of the recording they could make through a traditional turret-telephone system. That’s not the case for Cloud9's digital technology. Where a traditional turret solution might be 4 kilohertz, a little better quality than a traditional telephone conversation, it is a far cry from the 16 kHz captured through Cloud9’s system, according to Soto Sánchez. In addition to having crystal-clear recordings, Cloud9 benefited from the efforts of Quantiphi, an IT consultancy with offices in Boston and Mumbai that specializes in data science and machine learning. Cloud9 also stood on the shoulders of Google, which contributed its open source voice analytics software to the project.
“Trader voice comes with a unique set of problems for transcription. It’s full of slang and jargon, it includes overlapping speech and pitch variations, and the audio clips tend to be only a few seconds—all which make it incredibly difficult to transcribe,” said Asif Hasan, co-founder of Quantiphi. To solve these problems, Cloud9 and its partners used advanced deep learning technologies. The result: multiple trader voices can be transcribed "at unprecedented levels of accuracy within milliseconds," according to Hasan.
Cloud9 sees tremendous potential ahead with the newly accessible data. “You effectively have had a whole class of trading whose data has been unable to be accessed. We want to give people access to that, to use that data and gain better insights,” Soto Sánchez said.
