Voice analytics software developed by a Spanish company is helping to prevent fraud in the financial system.
Madrid-based Fonetic has developed software that allows financial services companies to monitor and assess telephone conversations involving their traders, to detect potential fraud, identify risk patterns and flag up market abuse or unwanted behaviour.
The company says its software – which has been used for five years by Spanish banks including BBVA and Santander – is up to nine times more effective than transcription-based analytics. Simon Richards, chief executive of Fonetic USA, said he hoped institutions in the City of London and Wall Street would soon deploy it too.
“If a regulator says, ‘I want all the conversation you’ve had about swap derivatives’ … to find that would take weeks. We can find them in 30 seconds. It gives [institutions] the ability to find conversations where key words have not even been mentioned.”
Fonetic’s software can recognise 79 languages, is designed for noisy environments, such as trading floors, and can identify individual traders by voice. It can also “find calls where there is a probability of lies taking place,” Richards added.
Although Fonetic’s financial services clients do not share with it how many times the software has successfully detected fraud or unwanted behaviour, Richards said it would be effective in preventing abuses such as the attempt in recent years by some banks to manipulate Libor, an important financial benchmark.
It can also help institutions comply with regulations such as the Dodd-Frank Act and the Markets in Financial Instruments Directive, which requires them to log specific trades.
Fonetic was founded in 2006 by engineer Juan Manuel Soto, now chief executive. The technology was developed in partnership with Genesys, a US group working in speech analytics, and Nuance Communications, a US speech recognition company.
Many of the 45 people employed by Fonetic in the Spanish capital are computer scientists, mathematicians and linguists drawn from academia.