In another example of linking trading data with communications, Fonetic, a Spanish company that specialises in speech processing, is able to provide trade reconstructions allowing someone in the compliance department to specify a particular trade and for the software to identify and group all the communications – including phone calls, instant messenger messages and text messages – relating to that trade.
The solution, which has been implemented by BBVA, Santander and others, can detect suspicious conversations even when traders are speaking in euphemisms, and it also takes into account the slang and vocabulary specific to the trading floor. Phrases such as ‘don’t talk about this too much’ or ‘this is between you and me’ are automatically flagged, as are promises of coffee or champagne.
Fonetic is also able to use linguistic models to detect all the endings of a particular word. For example, in detecting the use of the word ‘high’, the software would also scan for the words high, higher, highest and height, and a phrase such as ‘keeping it high for a few days’ would also raise an alert.
Juan Manuel Soto, CEO and founder of Fonetic, explains that a unique feature of the solution is that it is able to process speech directly from the audio, without transcribing the audio into text first. “Converting audio into text is a very poor process – you may only have 50% of what is said correctly transcribed,” he says.
Fonetic also has voice biometrics capabilities built into its solution, which Mr Soto sees as complementary to the company’s speech recognition software. The advantage of using biometrics, notes Mr Soto, is that a user can confirm that the counterparty is who they say they are. And, by analysing voiceprints, compliance teams can build a profile of a particular counterparty – even when they do not know who that person is – based on their voiceprint, and can bring up all the conversations that their traders have had with that person. The system can detect that it is the same person, explains Mr Soto, and can tell how many times the traders have spoken to them.
Big brother is watching
Mr Soto says that Fonetic’s solution is mostly focused on the analysis of the content of the conversations. At Nice, Mr Fraser says that the company has been discussing the use of biometrics with its clients, but for now its focus is on using speech analytics to review everything that is being discussed by traders and cross-referencing this with the surveillance of trading patterns.
“Most people are interested, at the moment, in understanding the information in the communications,” says Mr Fraser, adding that the use of biometrics will probably become more pervasive in the next year or so.
For further information you can log into “The Banker” clicking here.