Intelligent Voice Agent Improves Customer Service for Government Agencies

April 2018
Topics: Artificial Intelligence, Human Language Technology, Machine Learning, Natural Language Processing, Software Engineering
If smart machines could answer the typical questions asked by callers to government agencies, it could greatly reduce customer wait times. MITRE's solution does just that by interacting with customers, improving service, and streamlining communications.
MITRE employees interacting with iVAS

Callers to government agencies often hear messages like, "Thank you for calling. Your call is important to us. You are the 10th caller in the queue and your expected hold time is X amount of minutes."   

The more callers in the queue, the longer the "X" amount of minutes—and the higher a caller's frustration.

Although government agencies have web-based resources to assist customers, extended hold times at federal agency call centers persist. Exploring how government agencies can improve their interaction with the public, a team of MITRE researchers developed a voice agent that can effectively answer the typical questions asked by callers. Using it leaves agency staff free to answer the tough questions, which significantly reduces caller wait time.

Say "hello" to MITRE's Intelligent Voice Agent & Service, or iVAS. The interactive voice agent incorporates both natural-language understanding and artificial intelligence technologies to interpret a caller's spoken questions in context, understand the different ways people ask the same question, and asks the caller for specific information to clarify the question. iVAS then delivers precise, personalized answers.

AI and Natural Language Understanding: A Powerful Combination

Unable to find a commercial speech-recognition technology that could meet their needs, various government agencies began exploring their options. MITRE's experience in speech recognition, natural language learning, and artificial intelligence provided the perfect combination of expertise to develop a system that could be adapted for use across all government organizations.

MITRE's Jim Lockett, Qian Hu, and an interdisciplinary research team are researching and developing iVAS to meet the specific needs of government agencies—for example, it aims to work with all the commonly used voice assistants that people use at home.

"MITRE developed this technology to help government call centers provide meaningful, relevant, individualized answers that people can act on after making a call," says Hu, MITRE's chief scientist for speech technology and artificial intelligence.

(Video) MITRE has developed a prototype for an interactive voice agent that could help government agencies engage with customers and improve customer service.

Intelligent Voice Agent Effectively Solves a Problem

What makes iVAS different from commercial voice assistants like Amazon's Alexa, Apple's Siri, or Google Assistant? Hu explains that commercial voice assistants are search retrieval-based. They mainly deal with general and popular topics. In other words, they can't deliver the kind of concise, personalized, and relevant answers that callers to a government agency are seeking. 

"The iVAS system deals with domain-specific information and frequently asked questions for a particular agency," she says. "For example, if the system is being used by the IRS, questions and response are specific to tax-related topics."

Another feature that sets iVAS apart from commercial voice assistants is the system's interactivity. Natural language understanding allows iVAS to interpret and respond to the various ways a caller might ask the same question. Artificial intelligence enables the system to anticipate and ask additional questions, refining a caller's query to deliver the most relevant and personalized response. iVAS can also synthesize responses from callers so the system can adapt and learn.

Future Applications for Intelligent Voice Agent Technology

According to Hu, government feedback of iVAS has been very positive. "When we demonstrated the capabilities of the iVAS prototype to our sponsors, they were already thinking about how they could implement the technology."

The same underlying technology—natural language understanding and anticipatory question-and-answer capability—is also being used for government use of chatbots and text-to-speech recognition capabilities.

"There's a lot of potential for how this technology can be applied," says Lockett, a principal system engineer who specializes in information and agile systems. "For instance, the Department of Veterans Affairs could use it for depression screening and the IRS could help people check their refund status."

"When you have a multi-disciplinary team like ours working together, it creates the power to deliver greater capabilities," Hu says. "In this way we're better able to solve crucial problems for our sponsors, which means we're solving problems for the public."

—by Lisa Pacitto


Publication Search