fbpx

Conversational AI In Automotive

Meet Our Panelists

Shyamala Prayaga Ford

Shyamala Prayaga

Product Owner Digital Assistants @ Ford Motor Company

Shyamala Prayaga is a UX evangelist turned product owner with experience designing for mobile, web, desktop, and voice-based interfaces. As a part of AV Digital Experience Platform team at Ford, Shyamala uses these combined skills to direct product strategy from concept to launch for the AV Digital Assistant

Jeff Ferman Amazon

Jeffrey Ferman

Senior Solutions Architect @ Alexa Auto

Jeff has been involved in the automotive industry ever since growing up in Detroit and racing solar cars around the world. He leads the Alexa Auto Skills team, guiding partners as they build mobility-related Alexa skills. From remote car control skills to roadside assistance and parking, Jeff’s team is focused on delivering engaging voice experiences for customers on-the-go.

Robert Buchhardt Mercedes

Robert Bruchhardt

Staff Software Engineer @ Mercedes-Benz Research and Development North America

Robert has been working in the connected car area since 2014 and is a Software Engineer at Mercedes-Benz based in Sunnyvale, CA. He used to be a Product Owner for web frontend development before focusing on the connection of IoT and mobility use cases. On this quest, he is part of the #HeyMercedes engineering team which integrates voice with car interfaces to enhance vehicle UX. He is passionate about meaningful speech applications and finding the “everyday” use cases in the mobility sector.

Micheal Zogorsek Soundhound

Michael Zagoresk

VP, Product Marketing @ SoundHound

Michael Zagorsek is VP of Product Marketing at SoundHound Inc. where he focuses on strengthening the company’s presence in the voice market and expanding the reach of its core products: SoundHound, Hound, and Houndify. Throughout his career, he’s incorporated technology into marketing, whether it was digital advertising, launching websites, or innovative forms of human/computer interaction. Prior to SoundHound Inc., Michael served in leadership roles at Square, Leap Motion, and Apple.

Meet Our Moderator

Celene Osiecka 24 7 ai

Celene Osiecka

Director, Conversation Design @ [24]7.ai

Celene has been designing conversational interfaces using emerging technologies like chatbots, AI, natural language, speech recognition, and machine learning for the last fifteen years, deploying over 500 conversational interface deployments in the financial, telecommunications, travel, retail, and education industries. With a background in psychology and HCI, she currently leads a team of conversational designers across three countries that seeks to design innovative and ground-breaking conversational interfaces.

Event Q&A

Have any of you or your team segmented agents to be better tuned for situational jargon like a word that may mean one thing in the NLU you but have another meaning in a car?

Robert: Yes, [...] we have a lot of different sensor data, one of them being your navigation target, for example, or the state in which you are currently on. Our infotainment system, for example, supports profiles, so you can create a profile, for example for you or anyone. And then what we can do is [...] making use of that profile, to know who is driving the car.

Robert: We can then choose according to that, for example, to think about linking accounts to that profile. Or thinking about if you're heading to a specific place, then we probably want to search for coffee places along the route and not like someplace else at home or not even at your current location, but rather along your whole trip.

Jeff: I think that context is really important.

Jeff: We think a lot about your context of if you're in the car, and you say: "Turn on the lights", you don't want your house lights or bedroom lights to necessarily go on, you want the ones in the cabin; Or if you say set the temperature to a certain temperature, you're not necessarily referring to your living room.

Jeff: I think context is something that we always take into consideration. Another good example is if you're in the car and you say: "Play The Hunger Games", You may be talking about an audible book, you may be talking about a movie, you may be talking about a lot of different things. Recognizing that if you're in the car, playing audible is the right answer and not trying to start a movie and responding "Sorry, you don't have a screen", it would just be a frustrating experience.

Shyamala: Context is important. For example, a very simple thing, I come from India, we use centigrade, we don't use Fahrenheit, and here, I have to do the conversion when I have to change the temperature to x degree Fahrenheit.

Shyamala: So we tried to make our assistant smarter in the way like, even if I said: "make it 70 degrees centigrade", it would know 70 may not be the right temperature for centigrade, so maybe it should be Fahrenheit.
So, just trying to understand the intent of the user based on what they said, and try to smartly adjust.

What are some of the unique use cases that an auto specific assistant can offer that Google or Alexa can't?

Shyamala: One of the things I'm seriously looking at is the tour guide. You, sitting at home, you will not want to have a tour guide.

Shyamala: The other, which I think about is along the route. If there are any events, I want to know about booking tickets and stuff like that.
Of course, Alexa and others can also do it, but it's more relevant when you are in a new place and you want to get more information about those things.

Micheal: Shyamala talked about the user manual. If you have a question about the car: "How do I do this?", it makes sense to ask the car. People may be less inclined to ask Google or Alexa how the car itself works or "How to replace the wiper fluid?" or "What does this button mean?"

Micheal: Also as an extension of the service component, if you need to call a dealer, or if you have a customer support question, why not ask the system in the car saying: "Hey, I have a question about my subscription". The service component of the brand becomes integrated into the car in a really seamless way to a point where customer support technical support becomes synonymous with the assistant itself and that's something that the brand really is in the best position now to provide.

How automotive companies go about designing the persona, their system, and then how to validate that that persona resonates with drivers?

Robert: Mike already touched on the topic. If you think about the persona of a car, it's not only the voice assistant. It needs to tie in with the overall appearance and the overall perception of the brand.

Robert: It starts really with what happens to people if they see the logo, the front of the car, all the way to sitting in the driver's seat, and then talking to the assistant right. So this everything needs to tie together basically.

Robert: And then there are a lot of interesting questions. For example: "How formal should the assistant be?". Specifically, with Mercedes, we have a lot of different user groups. We have people like the stars, for example, who always sit in the back, and then you definitely need to accommodate this in another way, or maybe someone driving the "A class", being more tech. Getting all of this can be pretty hard.

Robert: And I think one of the keys there is personalization. The more we know about the user the more we can make use of their behavior and better adapt the personality of the systems.

How do you feel about wake words versus natural language understanding?

Micheal: We talk a lot about wake words at SoundHound because it's what unlocks the brand experience. It's a very human thing. So if I were to address you as Kai, then immediately I've developed the connection with you because you know I'm speaking with you.

Micheal: The wake word is a very important branded component, and that's one that we market pretty aggressively saying: "If you really want your branded voice experience, then you really need a wake word". You can do push to talk, and that's the default and that still can work the same way. You can still speak, but it feels a little bit more like dictation. Your brain just says: "I'm pushing a button. And now I must speak very simple words", whereas when I address you as a name, it inherently feels more conversational.

Micheal: It's just how we're wired, and then in the wake words the technology always lives embedded on the vehicle itself. The wake word is a powerful psychological component that I think needs to play a role.

Shyamala: Wake words are great. They reflect personality.

Shyamala: We are going one step ahead, where we are also experimenting with allowing you to choose a custom wake word like you name your car. You choose what you want. Then you are relating to the brand more because now you are just calling it by the name you call it, and you are able to connect with the assistant.

Robert: I totally agree with the wake word in the beginning. I'm also a huge fan of that. But what you can imagine and you already see that in a bunch of products out there is that for follow up things you don't necessarily need to say the wake word again. For example, if you have a timer running, you can just say: "stop" or these kinds of things. Another example, if you have one question and then immediately you ask another one without the need saying the wake word over and over again, because then at some point it gets weird.

Robert: That ties a little bit with the story from Mike before that. In the beginning, I would probably address people I would say: "Hey", "Hi". And then I would ask you something, but after that, the conversation is established, and then I wouldn't necessarily repeat your name all over again. So I think it makes sense to have the wake word in the beginning, but after you have established that there's a conversation ongoing, you can be smart about follow-ups.

Curated Resources By Voice Tech Global

Is your car’s infotainment system a serious distraction when using voice?

Source

This study by the TLR outlines some of the advantages of using voice over touchscreen for drivers.

cai-automotive-tlr-interacting-with-in-car-assistant

Adaptative In Car Assistant For Autonomous And Manual Driving

Source

This integration by Elektrobit for Alexa shows the extended proactive assistant and how it can adapt depending on the situation.

Voice User Interface Modern Guide

Source

In this guide, SoundHound with a lot of partners share a set of best practices acquired through experience with customers across many situations.

cai-automotive-Guide-to-Building-a-Better-VUI-2019-from-Houndify

Trends And Future Of Voice In the Car

Source

In this report, you will get a glimpse of what is the state of the union for Voice in the car as well as a map of the current key players.

cai-automotive-voice-on-the-go-report-capgemini

In Depth Design Of In Car Assistant Personality

In this paper, you can see the in-depth study of BMW and european schools on in car assistant personality

cai-automotive-bmw-real-world-driving-study

Consumer Report

Source

Webinar: Findings of the report presented by Bret Kinsella and Sanjay Dhawan.

Voicebot.ai consumer adoption report for in car voice assistant.

cai-automotive-in_car_voice_assistant_consumer_adoption_report_2020_voicebot

table of content

×