AI-driven 911 translation saves first responders time, money
David Degner/Getty Images
A Louisiana parish has turned to the technology to help translate emergency calls from a growing number of non-English speakers and amid a staffing shortfall.
Calls come into 911 when people are at the “worst point in their lives,” said Karl Fasold, interim executive director of the Orleans Parish Communications District in Louisiana. When those callers are “word vomiting” in another language, it can be tricky for dispatchers who aren’t familiar with the language to help them.
That’s why several years ago the parish turned to artificial intelligence, adopting an AI-driven one-way service that automatically translated and transcribed what the caller said about an incident. Now, the parish is among the first to use a live two-way audio translation service that it hopes will get emergency personnel out the door much faster.
The Orleans Parish Communications District, which includes the city of New Orleans, has seen the number of non-English speakers grow over the past two decades. An influx of Spanish speakers moved to the city after Hurricane Katrina to help with cleanup and recovery, and the city is home to a large contingent of Vietnamese residents. That’s in addition to half a dozen other languages, including the French and Cajun French traditionally spoken in the state.
The variety of languages spoken creates a challenge for 911 call takers, and that’s where AI comes in.
Under the old system, dispatchers would receive the initial details of the call via the one-way translation and transcription service. The caller and dispatcher would then be connected to a three-way phone call with a commercial translation service, which would help further facilitate the conversation so dispatchers could gather more information.
But the need to engage the third-party translation service sometimes meant delays of up to a minute as dispatchers worked to understand the issue, meaning crucial time was lost in an emergency.
The new live two-way translation services, which allow the dispatcher to hear the caller’s original audio and a translation, are up to 70% faster, according to parish officials. They also help mitigate the ongoing staffing struggles faced by 911 call centers nationwide. Needing to be fluent in multiple languages is something Fasold said is “not going to happen.”
“It's hard enough, given the current job market and the nationwide shortage of personnel in our industry, for us to keep sufficient people in the chairs to answer the calls at all,” he said.
The new two-way service is an upgrade to the agency’s old system, which the emergency communications firm Carbyne rolled out last year. Fasold said the district is an “early adopter” of that new feature, and said it will save time and money compared to using the old third-party translation service.
Developers are still training the AI technology, especially when it comes to making it familiar with the various street names people may be calling from. Many French or Creole-derived names are not easy to pick up on a transcription or translation service. Fasold said the need to ensure addresses are correct for first responders shows the need for people to work with AI, not be replaced by it.
Already, the system has been trained on previous calls to the parish from non-English speakers, and Carbyne had a native speaker listening to what the caller was saying to validate the AI system’s interpretation.
“That's why we call it computer-assisted dispatch, not computerized dispatch,” he said. “We're always going to need humans to put context into it, to understand the emotions involved in the call.”
AI will continue to be a crucial part of helping first responders and 911 dispatchers, said Alex Gruber, Carbyne’s vice president of product. In addition to its language and translation services, the company is looking into using AI to help manage an influx of calls around a specific incident. When a lot of people call in at once about the same incident, they can overwhelm call centers and prevent other time-sensitive calls from being answered quickly.
Dispatchers could plot the incident on a map, then see if there are other calls from that area and send those calls to an AI bot to ask if they are calling about the same incident or something else.
That could help triage calls to overwhelmed 911 centers, and the AI bot does not preclude people in the area from sharing more information about the big event. “We want to make sure you always get to the human when and as you need it,” Gruber said.
Fasold said he and his colleagues in Orleans Parish “have not figured out a downside” to the technology or any new potential impacts it could have for 911 call centers and first responders.
NEXT STORY: California suspends Cruise robotaxis, and questions mount