How to Solve the Voice Assistant Accent Gap
Want to confuse your voice assistant? Have a bunch of Bostonians ask it questions.
Voice assistants have an accent problem. As reported by The Washington Post, Google Home and Amazon Alexa devices “showed notable disparities in how people from different parts of the U.S. are understood.”
The Post’s investigation found that accents all over the U.S. didn’t always receive accurate responses from their voice assistants. The biggest issues were discovered by people with non-native accents. According to The Post, “People who spoke Spanish as a first language, for instance, were understood 6 percent less often than people who grew up around California or Washington, where the tech giants are based.”
What does this mean for the prospect of voice apps? For starters, market growth could be stunted if people feel they aren’t being represented or understood. One reason is social as people who are being misunderstood feel isolated, making voice assistants almost unusable for them. A second, more practical reason is that Voice is a platform that is fighting to be a first class citizen alongside Mobile. If users find that Voice is a difficult way to engage, they will be hard pressed to move off very mature mobile applications that they know and love. Think Starbucks; it has an incredibly popular and highly rated Mobile application, so what is the user’s motivation for moving off of the app if Voice isn’t proving to be an easier, reliable alternative?
Both Google and Amazon are not distancing themselves from the problem and have been taking aggressive steps to fix it. According to the article, “The companies said they are devoting resources to train and test the systems on new languages and accents, including creating games to encourage more speech from voices in different dialects.”
At Applause, and due to the nature of what we do, we are seeing an increased demand to help our existing and new customers solve these problems as they take on the challenges of Voice interaction as a core offering, not just a nice-to-have. What they are realizing is that lab-based, in-house testing simply cannot scale to test for all the variables that must be accounted for to uncover accent barriers, including languages, dialects, genders, and devices. It is becoming clear that real world testing with a variety of real users is the best way to uncover such issues.
A perfect example is the Auto Industry, where Applause works with seven of the 10 largest auto manufacturers. Global in nature and with sophisticated electronics already, Voice is a natural application for the auto industry. We saw first hand the challenges of the accent gap in action when testing for one of these leading companies.
This company invested heavily into integrating voice recognition technology into its infotainment systems. The company tested the integration extensively in the lab before releasing into real automobiles, and then continued to test in the lab with new generations of the vehicles, the voice assistant, and mobile devices.
Through internal, real-time monitoring, the car manufacturer saw customer utilization was fairly high. However, customer sentiment was not keeping up with the adoption rate. This poor sentiment was impacting drivers’ perception of the car manufacturer, its vehicles, and the voice assistant.
Leveraging the Applause community and providing for a diversity of dialogs that could not be scaled internally, the customer seamlessly scaled its dialog verification testing, and was able to identify where this negative sentiment was starting. We found the voice assistant the infotainment systems integrated with struggled to understand and interpret commands spoken by users with specific types of Southern and Midwestern accents. Considering this was just a subset within North America only, the feedback helped the company adjust its strategy to provide a better all-around driving experience and ensure it had a plan in place to support its global rollout.
Voice technology is clearly here to stay and will continue to rapidly grow as it is already a staple of many households. It is important to remember the technology is still a work in progress, with a number of new applications being tested and discovered.
Consumers will ultimately determine the winners and losers in Voice, just like they have with Mobile. It is encouraging that companies recognize improvements need to be made and are actively testing the experiences they deliver to users.