Want a covered parking spot at a public garage that takes cash and has a charger for your electric vehicle? Today’s voice-recognition systems would be overwhelmed by such a detailed query.
But Nuance, the voice and language technology and services supplier, thinks it can help, showcasing its cloud-based contextual reasoning framework, leveraging the natural language understanding capability of artificial intelligence.
Contextual reasoning leveraging AI works by accessing a large repository of knowledge, both embedded and in the cloud, in combination with driver preferences, location and situational context, and vehicle-sensor data, Nuance says.
“For us, what artificial intelligence means in the car context is the ability of the driver to say relatively complicated queries, and the system is able to understand that and offer sensible solutions,” Slawek Jarosz, senior manager-automotive applications for Nuance, says at Nuance’s annual Detroit auto forum.
AI also can help create “reasoning systems to fill in the blanks that occur when people talk. There’s a lot of assumptions made when you say something,” Charles Ortiz, director-Laboratory for Artificial Intelligence and Natural Language Processing at Nuance, tells WardsAuto.
Jarosz, whose team integrates voice-recognition-engine technology, cloud technology and writes dialogues, demonstrates the type of complicated query he says contextual reasoning makes possible.
“Find parking near the Michigan Central Station from 4 p.m. to 8 p.m. that takes cash and is for visitors,” Jarosz asks a prototype Nuance concierge system, which uses the supplier’s Dragon Drive connected-car framework.
The system responds, “It seems none of the places is for visitors and customers. The best match is Motor City Casino Hotel about half-a-mile away from there. Want this one?”
Contextual reasoning also can handle disjunction or the ‘or’ that results when more than one option is feasible in a given query. Jarosz asks the system to find valet or covered parking near the Detroit Opera House. He cycles through the results by saying “next” and settling on the fourth result, the farthest away but which has valet parking and is not covered.
On filling in the blanks, contextual reasoning comes into play if a driver asks “for a restaurant tonight.” The word “tonight” indicates a time element and will trigger system responses asking the driver if he would like to make a reservation and how many people will be in the dining party.
If a driver asks the system for a particular type of restaurant, for instance, one serving Italian cuisine, then soon after he changes his mind and asks for modern American food, the system will discount results pertaining to the earlier request.
By accessing sensor data from the vehicle’s bus, the concierge system is smart enough to tailor results to the current situation. For instance, it can know when an electric vehicle has a sufficient state-of-charge in its battery, resulting in a wider array of parking choices including garages that don’t have charging stations.
Contextual reasoning also builds on driver behavior, with Jarosz noting he has asked for parking garages that take cash so often that his results are biased to cash-only garages without specifically having to request them.
The demonstration Jarosz gives uses data from a variety of third-party vendors, including Parkopedia, Opisnavx and HERE, but automakers likely would select their own preferred vendors for parking, gas-station and mapping data, based on which provider is strongest in the region a vehicle will be sold.
While the contextual reasoning framework strictly is a prototype, Jarosz notes Nuance has picked up some customers for its voice-biometrics technology, making possible a concierge system recognizing users by their voice. He cannot say when it will arrive in market.
