A Considerate and Responsive Onboard Assistant for Cars

Researchers at Bielefeld University have developed a responsive voice-dialogue system

With the help of interactive voice-dialogue systems in cars, drivers can control navigation instruments or the radio with their voice. This technology is meant to prevent distracted driving. However, the opposite is often the case. Voice-dialogue systems frequently continue speaking to drivers when they have to navigate difficult traffic situations. In order to stop these electronic assistants from bothering drivers in the future, linguists at Bielefeld University have developed a new voice-dialogue system that stops talking when it recognizes that the driver is performing critical maneuvers.  

The research team tested the voice-dialogue system capable of “holding back” in the driving simulator, which showed that the team’s new system is virtually distraction-free. Photo: CITEC“In critical situations, such as overtaking another car, our capacity to concentrate is not enough to manage additional tasks,” explained linguistics researcher Professor Dr. David Schlangen of Bielefeld University. “The driver is forced to multitask, which makes both tasks – driving and listening – more difficult for him.” Dr. Schlangen referred to new studies from the University of Utah (USA) and the Foundation for Traffic Safety in Washington (USA). These studies confirm that voice-control systems in cars make driving less safe because they are often quite complex or prone to making errors.

The new voice-dialogue system from Schlangen’s research group “Applied Computational Linguistics” is designed to eliminate these problems. The group belongs to the Faculty of Linguistics and Literary Studies and is part of the CITEC Cluster of Excellence. Unlike previous voice-dialogue systems, this new system can respond to new traffic situations as they develop and hold back at critical times.

In an experiment with 17 study participants, Dr. Schlangen and his team tested the system in a driving simulator. Their experiment investigated how well their system performed in comparison to a conventional one 

Prof. Dr. David Schlangen, of Bielefeld University, develops computer systems that can understand and use language like humans. Photo: CITECDuring the experiment, the study participants had to drive on a simulated highway where they controlled the steering wheel and the gas pedal. A light signal indicated when the test subjects were supposed to overtake another car. During the driving simulation, the voice-dialogue system – either Dr. Schlangen’s new system, or a device working in the  conventional way  – checked in with the driver. The system alerted the driver to his or her pending appointments (e.g. “You have a lunch scheduled at 1pm with Peter. You are meeting in the cafeteria”). The conventional voice-dialogue system then continued to speak while the driver was in the midst of overtaking the car. The new system, on the other hand, stopped speaking as soon as the driver began to overtake the car. It spoke again only once the maneuver was complete. In order to sound more natural upon resuming, the system repeated part of what it had already said.

“We could prove that our system, in comparison to a conventional system, is far less distracting. It has a significantly less negative influence on the driver’s performance,” said Schlangen. “Thanks to our new technology, the driver is able to drive more precisely and, at least in the simulation, the danger of having an accident dropped. Conventional voice-dialogue systems in cars actually impair the driver’s performance as much as making a phone call while driving.”
In addition to measuring the impact of these systems on driver performance, the researchers also tested whether a study participant was able to correctly remember the appointments given by each respective voice-dialogue system. “Here we saw that most of the participants could remember the appointments if they were told by our system,” said Dr. Spyridon Kousidis, who was one of the leaders of the study. “In contrast, study participants were less frequently able to remember the appointments when they were told by the conventional system that could not interrupt itself.”

The researchers compared their system not only with conventional systems; they also determined how easy driving was for study participants when they was no voice-dialogue system present at all. The results: “Our system works unobtrusively. Drivers are no more distracted by our system than they would be if there were no voice-dialogue system present in the car,” said Dr. Schlangen.

Contact:
Prof. Dr. David Schlangen, Universität Bielefeld
Faculty of Linguistics and Literary Studies
Telephone: +49 521 106-67323
Email: david.schlangen@uni-bielefeld.de