"How to build cool & useful voice commerce applications (such as devices like Alexa & Google Home)" by Victoria Livschitz, founder of Grid Dynamics (Conversational Interaction Conference 2.11.20)

•

0 likes•138 views

In this talk, we describe how to build conversational e-commerce applications for the growing market of voice-powered AI devices using Dialog Flow. This talk demonstrates the capabilities of "Flower Genie," a teaching-oriented chatbot that can recommend a bouquet for any occasion, then take an order and deliver the flower arrangement via Alexa. We present the overall architecture of a voice application developed with Dialog Flow, including dialog management and NLU, and then discuss the finer points of testing and publishing a voice application for Alexa.

Technology

1
How to build cool & useful voice
commerce applications
(such as Alexa & Google Home)
Victoria Livschitz, Founder
February 2020

2
Today’s talk
・ At Grid Dynamics, we’ve been working on conversational systems since 2017
・ Particular focus on voice commerce: selling products & services via Alexa/Google Home/etc.
・ So, what does it take to write a conversational application?
Let’s take a close look at the “Flower Genie” and “Data
Genie”: our “Petshops of Conversational AI”

3
So, you want to order some ﬂowers for you mom?

5
But it’s not all roses. Typical gotchas and hiccups

8
How about an AI recommendation system for a camera?
・ Hint: a lot harder than flowers!
・ 1,000s of products; wide range on categories, from cheap point-n-shoot to professional
・ Technical product with wildly diﬀerent features
・ Some customers know exactly what they want (hobbists, pros); others - nothing at all
Help me choose the right camera. How hard is it?
What else do we want from the dialog?
・ Know the diﬀerence between advisory vs. order-taking. Adopt the dialog accordingly
・ Determine customer’s knowledge level. Adopt the dialog accordingly
・ Graceful switch between “leading the witness” & provide useful information.
・ Deal with “I dunno”

9
Live demo!
Scenario 1: “I want a camera for hiking”, “I want a camera for travels”
Scenario 2: Expert that knows exactly what he wants
・ Step 1: quickly determine that I don’t know much about cameras
・ Step 2: gently find out what I need camera for, then lead me through selection
・ Step 3: ask and answer reveland (to the purpose) questions
・ Step 4: close the deal
・ Step 1: quickly determine that I already know a lot about what I want.
・ Step 2: provide direct, complete, factual information. Follow, not lead.
・ Step 3: close the deal

10
Conclusion
・ Availability of high-quality AI models is spreading =>
・ The price & complexity of conversational applications is rapidly coming down =>
・ Best practices in design, testing and certification of conversational apps are emerging =>
・ This is already practical to create a wide range of useful eCommerce voice / visual apps today

11
Understanding customer queries using Deep Learning

12
ML bag of tricks
1.
2. 3.
Transfer learning

13
Design cycle: yep, it’s a cycle. Particularly in AI systems.
If you don’t understand your customer’s behavior, your customers will not understand your AI

15
Finally, testing and certiﬁcation. There is a lot to it.

16
To learn more
・ Detailed blog post that spills all the beans:
https://blog.griddynamics.com/how-we-built-a-conversational-ai-for-ordering-flowers/
・ More about conversational AI development services:
https://www.griddynamics.com/technologies/ai/voice-application-development-services
・ Write to us!
vlivschitz@griddynamics.com
info@griddynamics.com

Similar to "How to build cool & useful voice commerce applications (such as devices like Alexa & Google Home)" by Victoria Livschitz, founder of Grid Dynamics (Conversational Interaction Conference 2.11.20)

Usage data allows PMs, the product team, and the whole organization to make better decisions. Good usage intelligence gives you the ability to be smarter, more active, more decisive, nimbler, and to minimize risk. But what if you don't have that data - such as before you have users? Or, what if the right decision seems to fly in the face of the data you have? Or, what if your product offers more than just the standard features? To get deeper into these questions, Nils Davis asks, "What is the most interesting thing about Instagram?" (Because who doesn't like a product that Facebook paid $1 billion for when it had fewer than 50 employees and no revenue?) Nils will use the example of Instagram’s Filters to talk about how putting prebuilt knowledge in your product can change the way your product is used for the better - putting you in the company of most market-leading products. Finally, he’ll tie it all together by explaining how the way you interpret and use usage data can impact the way your tell your product’s story, and ultimately, how your users use your product.

Analytics in Action: What Users Want: How and Why to Build Knowledge into You...

Hannah Flynn

Internet of Things meets UX #Flipthinking - Globant San Francisco 2014

Globant

PhoenixGMN All About Apps - Idea to Launch

PhoenixGMN WebAndAppsDevelopment

How to find & pick a tech agency

Koombea

From machine learning and augmented reality to chatbots and voice-activated search, there’s a lot of buzz around how to use new technology to improve your marketing efforts. But sometimes, it’s hard to know whether or not they’re actually worth the investment. In our webinar led by E3 Senior Marketing Technology Manager Grady Neff, we cover some of the latest marketing technologies, how marketers are using these technologies to move their brands forward, and how to evaluate solutions for your own company.

Emerging Marketing Technologies Webinar

Element Three

Reshaping Your Business With Web2.0

Billy Cripe

How to avoid 6 deadly mistakes when building a digital product 2018

inFullMobile

What every developer can learn from startups

Oleg Podsechin

IoT Product Design and Prototyping

Dr. Shivananda Koteshwar

Eye see overview general capabilities

EyeSee Research

Rapid Prototyping & Customer Development

johnwlong

2019 InfoSec Buyer's Guide

Adrian Sanabria

Using Product Box to Build the Complete Developer

Luke Hohmann

Better Service Management with Artificial Intelligence

TOPdesk

The Heek Product Cycle

Heek Team

It career paths

brilliantweb

Better Service Management with AI

TOPdesk

Iterating on Developer Marketing with Metrics

Jonathan Gottfried

As technologists, we love to build things. And we sometimes forget that our customers (or potential customers) don’t care about what we’re building. They care about what they’re building, doing, or feeling. In this talk, we’ll explore methodologies that help us continually focus on our customers’ needs, building just enough to learn and iterate towards their desired outcomes. Coming away from this, you’ll have a few more tools in the toolbox for your lean startup.

MVP to MLP - Minimum Lovable Product

Jake Levirne

How to do an effective seller presentation on a tablet

Better Homes and Gardens Real Estate

Similar to "How to build cool & useful voice commerce applications (such as devices like Alexa & Google Home)" by Victoria Livschitz, founder of Grid Dynamics (Conversational Interaction Conference 2.11.20) (20)

Analytics in Action: What Users Want: How and Why to Build Knowledge into You...

Internet of Things meets UX #Flipthinking - Globant San Francisco 2014

PhoenixGMN All About Apps - Idea to Launch

How to find & pick a tech agency

Emerging Marketing Technologies Webinar

Reshaping Your Business With Web2.0

How to avoid 6 deadly mistakes when building a digital product 2018

What every developer can learn from startups

IoT Product Design and Prototyping

Eye see overview general capabilities

Rapid Prototyping & Customer Development

2019 InfoSec Buyer's Guide

Using Product Box to Build the Complete Developer

Better Service Management with Artificial Intelligence

The Heek Product Cycle

It career paths

Better Service Management with AI

Iterating on Developer Marketing with Metrics

MVP to MLP - Minimum Lovable Product

How to do an effective seller presentation on a tablet

More from Grid Dynamics

Industry-leading brands have adapted to the ever-changing business environment and consumer expectations significantly faster than the competition. The key driving factor behind this speed-to-market is the adoption of intelligent cloud-native digital commerce architectures. In our talk, we will describe the best practices and case studies of achieving business agility that is imperative to excel in the post-pandemic era.

Are you keeping up with your customer

Grid Dynamics

The quality of business decisions, machine learning insights, and executive reports depend on the quality and integrity of the underlying data. There are many ways that data can get corrupted in an analytical data platform from de-synchronization with the system-of-record to defects in data pipelines. We will show how to detect and prevent data corruption with automation, open-source tools, and machine learning.

"Implementing data quality automation with open source stack" - Max Martynov,...

Grid Dynamics

Dynamic Talks Portland: The use of AI in many industries has revolutionized operations and efficiency. In healthcare, the progress is just beginning. Despite the promise of AI, why has the development lagged other industries? What issues are unique to healthcare that create challenges for common approaches? How can data scientists overcome these challenges and deliver on the promise of using data to reach multiple goals of improved quality, decreased cost, and greater patient satisfaction?

"Challenges for AI in Healthcare" - Peter Graven Ph.D

Grid Dynamics

The global HIV pandemic continues, particularly in Sub-Saharan Africa. By 2025, 40 million people will be living with HIV. The global cost of the pandemic is in the hundreds of billions of dollars. Better treatment means more people are living longer and costs will increase. The use of existing and emerging technologies is rare. Research institutions don’t share data. Data that drive US HIV policy in 2019 are from 2017 because of the time it takes for the CDC and NIH to combine data. The are many opportunities for big data, ML and AI to have a broader and continued impact on the HIV crisis. The use of these technologies can identify new avenues of research and help prioritize and focus efforts. We are starting to see these technologies used more and more. Several case studies will be presented. For example, advances in HIV vaccine research by Dr. David Heckerman; research at UCLA and Georgetown University looking at how social media can be used for tracking and predicting the spread of the epidemic; and work by researchers to improve care utilization in South Carolina. The opportunities for commercial and non-profit ventures to apply existing and emerging technologies like big data, ML and AI are countless. Tech4HIV is an organization working to drive these efforts into the tech sector and provides opportunities and resources for engagement.

Dynamic Talks: "Applications of Big Data, Machine Learning and Artificial Int...

Grid Dynamics

Retail industry has gone through a significant transformation in the last decade; caused by the change in consumer behaviour and expectations which have been driven by the convenience and low cost offered by Amazon. There are clear signs of a similar disruption in the financial services industry now, driven by new entrants who are changing consumer expectations. In this talk, Rahul looks at the evolution of retailers and draws conclusions about the path players in the financial services have to take to compete and provide customer experiences that will be necessary to compete.

Dynamic Talks: "Digital Transformation in Banking & Financial Services… a per...

Grid Dynamics

Dynamic Talks: "Data Strategy as a Conduit for Data Maturity and Monetization...

Grid Dynamics

Apache Spark is a general-purpose big data execution engine. You can work with different data sources with the same set of API in both batch and streaming mode. Such flexibility is great if you are experienced Spark developer solving a complicated data engineering problem, which might include ML or streaming. In Airbnb, 95% of all data pipelines are daily batch jobs, which read from Hive tables and write to Hive tables. For such jobs, you would like to trade some flexibility for more extensive functionality around writing to Hive or multiple days processing orchestration. Another advantage of reducing flexibility is creating "best practices", which can be followed by less experienced data engineers. In Airbnb, we've created a framework called "Sputnik," which tries to address these issues. In this talk, I'll show the typical boilerplate code, which Sputnik tries to reduce and concepts it introduces to simplify pipeline development.

Dynamics Talks: "Writing Spark Pipelines with Less Boilerplate Code" - Egor P...

Grid Dynamics

"Trends in Building Advanced Analytics Platform for Large Enterprises" - Atul...

Grid Dynamics

For businesses large and small, cloud-based software platforms are breaking down data siloes between teams, adding new capabilities, and boosting efficiency. While not known as early adopters, there's a compelling case for public safety to do the same. In this session, we'll explore how a comprehensive records and evidence platform can simplify operations and streamline data management and access, securely and responsibly. We'll also highlight the role of artificial intelligence, and how it can help agencies respond more quickly and improve outcomes.

The New Era of Public Safety Records Management: Dynamic talks Chicago 9/24/2019

Grid Dynamics

Dynamic Talks: "Implementing data quality automation with open source stack" ...

Grid Dynamics

Dynamic talks Seattle: Artificial Intelligence (AI) and data are foundational to ideation of new business models that bring growth and efficiencies. This is in action in pharmaceutical supply chain for reducing costs, increasing volumes, and optimizing the contracts with suppliers. The implementation involves process and tools that make data usable, and overcome the challenges of culture, ethics, data scalability, compute and engineering, and. Learn about data collection, data management, and metadata management tools implementation and modern data architecture to support them. Discuss machine learning algorithms for growth and efficiency scenarios.

"Implementing AI for New Business Models and Efficiencies" - Parag Shrivastav...

Grid Dynamics

Dynamic Talks Seattle: Patient No-shows and Late Cancelations is estimated to cost around $150B across the nation. Providence St. Joseph Health, Healthcare Intelligence Group has been working on this pain point for the past couple of years and has developed a software solution leveraging a predictive engine to identify high risk patients at risk of No-show and Late Cancellation on a daily basis. More than 25 clinics have been using this solution for about 2 years now to optimize their reminder call strategy and to have a meaningful reduction of No-shows and Late Cancellations. In addition, a call center pilot was designed recently to standardize this process end-end. Pilot results will be evaluated to prove the impact at scale.

Reducing No-shows and Late Cancelations in Healthcare Enterprise" - Shervin M...

Grid Dynamics

In this talk, we will discuss automatic decision-making and AI techniques for customer relationship management. First, we will present a methodology that helps to develop highly automated promotion and loyalty management systems. Next, we will walk through practical examples of how advanced customer and content signals can be generated using predictive models, and how optimization and reinforcement learning techniques can be used for targeting, budgeting, and pricing decisions. This talk is for Data Scientists, Product Owners, and Software Engineers involved in marketing operations or development of marketing automation software and interested in ML-based decision automation techniques.

Customer intelligence: a Machine Learning Approach: Dynamic talks Atlanta 8/2...

Grid Dynamics

Dynamic talks SF: So you have heard all the hype around how Machine Learning is going to change the world. But within your business context, where do you start? How do you get leadership buy-in for investment? And how and when you start scaling your ML Services? In this session, you will walk away with an actionable framework to bootstrap and scale a machine learning services team. You will see the framework in action through an actual 0 to 1 product journey involving deep learning where we delivered value in record speed in-spite of not having a dataset when we started. You will get practical tips on how to make decisions about when and how to scale your capability to scale ML Services and platform. Some of the tips are pretty counterintuitive and revealed themselves with our experience of productizing ML services over the last 5+ years. (Using a diverse range of technologies - Vision, Language, Graph, Anomaly Detection, Search Relevance, Personalization)

"ML Services - How do you begin and when do you start scaling?" - Madhura Dud...

Grid Dynamics

Dynamic Talks SF: Recommendation systems are all around us. E-commerce companies like Amazon recommend products we are likely to buy based on our browsing behavior. Netflix suggests what shows we should watch based on our binging habits. Spotify builds a personalized playlist we would enjoy listening to, based on their understanding of what musical genre we are into. In this talk, we will explore recent advances in the area of product recommendations in both research and practice. We will see how machine learning, design thinking and solid data engineering principles are combined to create an engaging customer experience that positively impacts the bottom line. We will look at how we use various deep learning architectures to obtain image and text embeddings that supplement user and product based features to generate product recommendations that align closely with a consumer’s aesthetic preferences. The talk would be of interest to data scientists, data engineers, product managers, UX designers and anyone interested in machine learning.

Realtime Contextual Product Recommendations…that scale and generate revenue -...

Grid Dynamics

In this talk, we will discuss automatic decision-making and AI techniques for customer relationship management. First, we will present a methodology that helps to develop highly automated promotion and loyalty management systems. Next, we will walk through practical examples of how predictive models can be used to characterize customer intent, and how optimization and reinforcement learning techniques can be used to build next best action models that incorporate targeting, budgeting, and pricing decisions.

Decision Automation in Marketing Systems using Reinforcement Learning: Dynami...

Grid Dynamics

When migrating to a cloud and microservices architecture, companies need to invest in foundational capabilities, such as a microservices platform, continuous delivery, and an immutable infrastructure. In this talk, we will discuss our experience implementing these capabilities on the enterprise-scale with Google Cloud, Kubernetes, Istio, Envoy, Spinnaker, and Hashicorp stack. We will also discuss best practices of onboarding the cloud to facilitate DevOps, SRE without sacrificing quality or control.

Best practices for enterprise-grade microservices implementations with Google...

Grid Dynamics

Attribution Modelling 101: Credit Where Credit is Due!: Dynamic talks Seattle...

Grid Dynamics

In this talk, we will discuss how predictive modeling and reinforcement learning can be used to build advanced price management systems that unlock the potential of dynamic and personalized pricing. We will present price optimization methods for a number of use cases including introductory pricing, promotion calendars, replenishable and seasonal products, targeted offers, and flash sales. We will also review case studies that demonstrate how these methods were applied in practice and how algorithmic price management components were fitted into pricing strategies.

Building an algorithmic price management system using ML: Dynamic talks Seatt...

Grid Dynamics

Customer intelligence: a machine learning approach- Dynamic talks Dallas Q2

Grid Dynamics

More from Grid Dynamics (20)