With coronavirus taking a significant chunk out of physical retail and services, the use of call centers and phone interaction with customers becomes even more important.
And it calls for advanced speech recognition technology that concentrates as much on what is said as it does on the voice inflection of those saying it.
For the past three years, London-based Speechmatics has offered its clients — generally companies that deploy call center services technology — with automated speech recognition that converts voice into text. The process essentially turns difficult speech-recognition processes, like someone speaking English in a foreign accent, into a "very consumable" product, said Ian Firth, vice president of products for Speechmatics.
"When people talk about speech recognition, they immediately think of Google and Alexa systems, which is all about uploading voices from people," Firth said. "But you can only take so much away from people's voices, as some human-to-human conversations with call centers can get very complex."
The challenge of speech recognition is heightened in a world where many people, forced to work and shop remotely, will interact with call centers for transactions they would otherwise have handled in person.
"With coronavirus now, it is even more important to have human-to-human available at call centers," Firth said. "But our technology works on both sides, so if companies want to handle more calls with less people, we can do that."
Fraudsters are likely to piggyback on this trend to take advantage of overwhelmed call center workers. Call centers have to be aware that fraudsters see those operations as potential gold mines, said Shirley Inscoe, senior analyst at Aite Group.
"More commonly, fraudsters are able to successfully impersonate legitimate customers and gain access to their online accounts, or order a new payment card," Inscoe said. "Of course, an insider in a contact center could certainly abet a fraud ring or leak data to such a ring."
As such, companies relying on call centers have to know that the person calling in is the person they say they are. Fraudsters using advanced technology can make that difficult for companies. Fighting back with exact text records of audio conversations helps companies comply with the General Data Protection Regulation as well as spot inconsistencies that trigger suspicions.
For Speechmatics, its role is making sure that all of the information shared during a call center conversation becomes data that companies can protect in the proper places or eliminate in compliance with GDPR. In addition, the company helps call center agents answer questions properly, also in compliance with GDPR.
This is important, too, in light of incidents like British Airways last year being fined €204 million — the largest GDPR penalty so far and one that signaled compliance as a top priority. Since July of 2018, data protection regulators from European countries have levied an estimated €458 million in fines.
"Our technology can listen to the calls and augment those agents, providing information that they can use in those conversations and to do it correctly," Firth said, referring to monitoring both human-to-human and human-to-machine calls. "They need to know what to say if someone is making a payment and provide the correct disclaimers regarding data."
To that end, Speechmatics last month expanded its partnership with recording and quality monitoring company Liquid Voice, which will further advance Speechmatics' automated speech recognition technology.
It is not uncommon for companies to parlay Speechmatics' automated speech recognition technology with other voice detection tools. But the conversion of voice to text is especially effective when determining what data to store and what to throw away.
GDPR tasks companies with keeping an eye on what occurs during audio customer service interactions. "What if the recording includes someone leaving a credit card number for payments?" Firth asked. "Under a normal process, it would be very hard to know it's there in the recording and strip that data out and store it somewhere."
With the automated speech recognition technology, the business would have a text version of that conversation, making all of the data more usable.
"You can make sure the data is going to the right places and you can take that card payment data out and put it into storage that would be PCI (Payment Card Industry data security standards) compliant," Firth added.
A key aspect of GDPR is that if a customer requests to know where a company stores certain data about an account or a transaction, the company has to have an answer.
"It is hard for me to say about individual businesses, but what people tell me is that the ability to analyze the data in this way helps them avoid stubbing their toe with GDPR," Firth said.
The technology can segment the conversation into different compliance categories, a job that would take humans a long time to process properly, Firth added.
Speechmatics was founded in 2006 and became commercial with its products in 2014, building a technology that can provide speech recognition to 31 languages. The U.K. and U.S. represent the company's largest markets, but it is serving markets across Europe and in Japan.
The addition of automated speech recognition to other security layers provides a way for companies to not only comply with GDPR standards, but to also watch for patterns in the conversations that could detect problems.
"The highly accurate transcription that Speechmatics provides allows us to better serve our customer while simultaneously minimizing the risk of data breaches, protecting finances and reputations in the process," said Simon Broaddbent, chief commercial officer at Liquid Voice.
