Extracting User Data from Speech Applications
And, of course, his business has been built on the use of the technology to help triage, expedite, and extract information related to emergency services.
One particularly interesting use case that Stevenson recalls was a Danish application where thousands of calls from people who thought they or someone else was having a heart attack were analyzed; the voice analysis engine was trained to detect through voice intonation or breathing whether someone was having a heart attack so those calls could be more quickly triaged.
AI is being used, says Stevenson, to pick up on intonation, intent, emotion, even background noises. But these types of applications are in their infancy, according to Eunjoon Cho, an engineer with four years of experience as part of a speech recognition team with a major technology player (he recently left to launch a start-up, Heya). His background is in speech recognition related to voice search—translating voice data to text and doing further interpretation to deliver the information users are looking for. That type of translation is fairly well developed, he says, but “there’s a totally different side where you look at cues”—whether the person is coughing, being sarcastic, etc. “I think it’s still in the very early stages,” he says. “Right now all of these smartphone devices are really dumb.”
Big Benefits to Intelligent Voice and Sentiment Extraction
“Most major corporations need to significantly cut costs and cut head counts so voice, AI, and machine learning can automate a lot of activity,” says Stevenson. Like Cho, though, he’s not seeing significant use right now. But, he adds, “I think they will become a much bigger play if we’re looking five years out.”
There are big benefits to be realized from technology that is able to do what Stevenson points to—efficiency in handling large volumes of customer or employee inquiries, improvements in service delivery, and deeper market knowledge gleaned through aggregated audio insights.
Call centers don’t represent the only obvious use case for this technology. There are efficiencies for salespeople as well, says Stevenson, particularly when these systems can be seamlessly integrated into tools like Salesforce. As with traders, they’re able to have sales calls transcribed seamlessly into the customer record.
In addition to becoming increasingly adept at getting callers to where they need to be quickly and seamlessly, automated systems that can be driven by voice prompts have another huge benefit from a service standpoint, says Ghaffar: the ability to provide access 24/7/365 across multiple time zones. This is clearly something that is particularly important to organizations that operate globally.
And, of course, the ability to analyze trends in caller requests and concerns has the potential to yield important and elusive market insights, as does the potential of using voice data in predictive ways.
All of these possibilities, of course, raise a variety of ethical and privacy-related issues.
Potential Ethical/Privacy Issues
When an Oregon woman claimed that her Amazon Echo Dots were secretly recording private conversations she had with her husband and, astoundingly, sending them to a random contact (an employee of her husband’s) almost 200 miles away, Amazon had some explaining to do. NBC reached out to Amazon for confirmation and an explanation, and its affiliate KIRO-TV received a statement from Amazon: “Amazon takes privacy very seriously. We investigated what happened and determined this was an extremely rare occurrence. We are taking steps to avoid this from happening in the future.”
It’s the kind of scary, albeit rare, scenario that can quickly spark privacy concerns and paranoia. And it has.
Consumers, consumer protection agencies, watchdogs, regulators, and others are understandably concerned. How do these concerns impact companies that may be capturing, aggregating, and using data—in this case voice data—for customer service, consumer transactions, and marketing?
The General Data Protection Regulation (GDPR) is one recent example of a far-reaching regulation, initiated in the European Union, that regulates personal data protection. It is legislation that applies to any company doing business (e.g., having contacts) in the Europe, and it is expected to spread.
Even the use of aggregated, not personally identifiable data can lead to a consumer outcry. Facebook’s Cambridge Analytica scandal may well represent the tip of the iceberg in terms of how aggregated personal data is already being mined for various reasons—some practical, some predatory. In this case, Facebook data from tens of millions of users was used to create, and sell, profiles of American voters to political campaigns, something the users had not agreed to.
Privacy issues are obviously a concern, and especially in areas of particular sensitivity, like private information numbers, Ghaffar acknowledges. The voice data extraction technology itself can represent at least part of a solution. For his clients, he says, it’s not unusual for employees to call in and say, “Hey, I’m Chris Smith and my Social Security number is…”; Kordinator is designed to parse out that data through sophisticated data dictionaries. “So, if we see there are nine digits, we parse that out, or we mask it before it goes to the agent,” Ghaffar says. His clients have fewer concerns than other companies may have, though, because they’re not using the data for marketing; and because employees have opted in to the system, they know they’re dealing with an automated system that is capturing their information, he says.
Stevenson’s perspective is that “generally, people seem to be more relaxed about data breaches—they happen every week and people don’t seem to respond.” Still, he says companies are taking the potential for breaches very seriously, with some pulling their data out of the cloud and back in-house, or using a hybrid model, making decisions based on the impact of privacy issues.
As large companies, especially those with global operations, begin to adopt this type of technology, new issues are likely to emerge, along with the complexities of navigating privacy concerns and regulations across geographies.
Despite exciting media reports about the emerging power of voice-driven data extraction and its potential, there are hurdles to be reached before its use becomes widespread.
Lin Pophal is a freelance business journalist and content marketer who writes for various business and trade publications. Pophal does content marketing for Fortune 500 companies, small businesses, and individuals on a wide range of subjects, from human resource management and employee relations to marketing, technology, healthcare industry trends, and more.
Companies still aren't doing enough to capture voice data, but advances in transcription, AI, and machine learning could change things