Conversational Toys – The Latest Trend in Speech Technology

Conversational toys seem to be all the rage. Two new entrants to the market are Mattel’s Hello Barbie and Elemental Path’s new CogniToy dinosaur. Before delving into the specifics of each toy, here’s a list of the primary features that both toys seem to share:

  • Speech recognition and natural language processing capability
  • Connection to the cloud
  • Ability to store basic information from previous conversations
  • Ability to offer personalized responses
  • Software that can evolve over time (get updates from the cloud server)
  • Activation of the toy’s listening mode by pressing a button

Barbie Chatbot Doll Powered by ToyTalk

Hello BarbieMattel is partnering with ToyTalk for the Hello Barbie doll. ToyTalk produces several popular mobile apps for children. ToyTalk also has speech recognition technology that’s specifically tuned to understand the higher register and more erratic speech patterns of children’s voices.

A Mattel spokesperson provided a brief demo of Hello Barbie at the recent Toy Fair 2015 in New York. Hello Barbie’s current conversational abilities are comparable to those of a chatbot that’s connected to Wikipedia or some other data source.

The talking Barbie can also store information about its conversational partners in the cloud, so that it can call on its memory to create more personalized responses. In the demo example, Hello Barbie remembers that its interlocutor enjoys being onstage. When the question of possible future jobs comes up, Hello Barbie uses this stored information to suggest a career such as dancer or politician, presumably since both jobs involve lots of time onstage.

According to the spokesperson, Barbie will have the ability to play simple conversational games, tell jokes and stories, and learn more about the person its talking to. The company hopes that offering this type of dynamic interaction with the doll will deepen the child’s relationship with it.

CogniToys by Elemental Path on Kickstarter

CogniToyElemental Path is currently blowing the roof off  the Kickstarter campaign to fund their CogniToys talking toy project. The last time I looked, they had raised over three times more than the $50K they were asking for, and there were still 23 days left in the campaign.

Elemental Path seems to have evolved from Majestyk Apps, which was one of the winners of the IBM Watson Mobile Developer Challenge.

Elemental Path is marketing their first production talking toy as both educational as well as entertaining. On their Kickstarter video, the cute dinosaur creature quizzes kids on simple math and counting exercises, answers science trivia questions, and tells knock knock jokes.

From the demo, it’s not completely clear how the CogniToy leverages the IBM Watson technology. Based on the Elemental Path website, the toy’s technology contains a dialogue engine that uses advanced language processing algorithms.

Concerns About Conversational Toys?

In an opinion piece for ComputerWorld.com, Mike Elgan writes about the potentially darker side of both Hello Barbie and CogniToys. Elgan’s main concern about the Barbie toy is that it pulls young children into a world of total surveillance. Everything children say to their Barbie is captured and sent to the cloud to be analyzed and stored on ToyTalk’s server. According to Elgan, ToyTalk will also email conversations to the parent.

In the case of CogniToys, Elgan expresses concern that the question-answering dinosaur teaches children that knowledge is stored in the cloud and served up by Artificial Intelligence. In the future, Elgan fears that children may not have to learn or experience on their own. Instead, they’ll just ask their intelligent assistant for the answer.

What’s the Market for Conversational Toys?

If the success of Elemental Path’s Kickstarter campaign is any indication, there might be a sizeable market for conversational toys. The fact that Mattel feels motivated to partner with ToyTalk is another sign that we need to take the trend seriously.

Will children be better off with conversational toys than without them? Only time will tell. It seems to me that there can be many positive outcomes to interactions with toys like Hello Barbie and CogniToys. It depends in large part on what producers “program” the toys to do, how interactive we can make them, and whether they will spark a child’s creativity and critical thinking capacity as opposed to stifling them.

CodeBaby Creates Virtual Guide for Colorado Healthcare Exchange

Derek Top of Opus Research recently wrote about intelligent assistants in the healthcare space. One of the technologies Top profiled was CodeBaby’s intelligent assistant Kyla, which supports Colorado’s health insurance marketplace. I gave Kyla a test drive to see how she works.

Codebaby KaylaCodeBaby’s Kyla differentiates itself from other self-service virtual assistants in that there is no text-based interface. In fact, you can’t ask Kyla specific questions. Though this structure sounds like it would be very constraining for the potential health insurance customer, I found that Kyla actually works quite well.

Kyla appears as an animated image of a young woman. The animation is pleasant and doesn’t mimic a human image enough to be creepy. Kyla is more like a guide than a question-answering bot. She pops up at the lower right of the healthcare connect screen and provides a pre-recorded message to welcome the user to the site and give them a quick overview of what the site is about. All of Kyla’s statements are pre-recorded. She has a human voice and her statements are made with natural intonation and tonality, which is a big plus over computer generated speech.

Once Kyla has finished introducing the user to the site (or to a new web page), a pop up appears with a selection of other topics that Kyla can address. Some examples of topics the user can choose are:

  • How long will it take me to enroll?
  • I want to learn more about financial options
  • What are the important deadlines I should know about?

When you select a topic, Kyla delivers her pre-recorded response. The text of her response isn’t displayed on the screen, so you must have your speakers turned on and you must be able to hear. You can start and stop Kyla’s recorded message and replay it as often as you like.

I think Kyla definitely works as a site guide and advisor. If a user has questions beyond those that have already been anticipated, Kyla won’t be able to assist. But she can refer the health insurance shopper to a webpage that helps them find a human broker. The intelligent assistant guide is a great fit for a website as complex and intimidating as a health insurance marketplace. It’ll be interesting to see if the intelligent assistant as guide gets expanded to other use cases

xDroid – An Intelligent Assistant for Social Search

xDroid is an intriguing personal intelligent assistant project that launched on Kickstarter a while back and that has only a few days left to meet its funding goal. The project was incubated at Columbia College Chicago’s Business & Entrepreneurship Department, so it comes with a solid pedigree.

I’m not sure why the app is called xDroid. It doesn’t seem to be tied to the Android operating system. The demo videos show the app working on an iPhone. But regardless of the name, the xDroid has a different and interesting take on the concept of what makes an effective personal intelligent assistant.

xDroidTo try and summarize the concept in a nutshell, xDroid’s main focus seems to be to connect you with things. Specifically, it aims to connect you more easily to people who are offering products or services that you are looking to buy. Likewise, it can also help you find customers for the things you hope to sell. It also connects you to others who have information that might be useful to you.

The app learns about you, your preferences, and your social network. If you need an illustrator for you latest book, xDroid will let others in your network know and it will search through the network to see if there’s an illustrator out there offering services. xDroid will facilitate a connection between you and the illustrator, or give you information about competing illustrators so that you can select the one you want.

Two aspects of xDroid stand out: the extremely futuristic and innovative user interface and the “social search” concept. Let’s start with the user interface. The animated screen captures on the Kickstarter page look like something off the dashboard of an alien spacecraft. The underlying concepts seem to be about showing connections between things: connections between you and others in your extended network, connections between your preferences and what others are seeking to buy and sell, connections between your friends and their current activities and interests.

The concept of social search isn’t new. Neither is “social selling.” That’s what LinkedIn is all about, I suppose. But the xDroid concept seems a bit more seamless. Based on the way the app is described, it appears that xDroid can anticipate what you want to buy and initiate searches in the background that span the web as well as all of the contacts in your extended network. The xDroid search engine pulls your friends and acquaintances into its search algorithm. It searches the real world, not just the web. Crowdsourcing replaces the drudgery of shopping. It’s a cool idea. Personally, I’d find the concept of social search even more intriguing if it wasn’t as much about buying and selling as about sharing ideas and learning from your social network. If you’re trying to solve a tough problem, xDroid could help quickly connect you to experts or others interested in helping you find a solution, for example.

I don’t know how much of the xDroid app is built out and how much is still in the concept phase. The Kickstarter campaign seems to have stalled, but I wish the team luck. It seems to me that they’re on to something.

Software Advice’s Report on Self-Service Channels

The team at Software Advice, an online consultancy for customer relationship management software, has published the results of a survey on customer self-service channels. The report contains interesting information on the effectiveness of a range of self-service technologies and how companies measure performance. Virtual assistants are included in the study.

SurveyTo obtain the results, Software Advice surveyed 170 professionals within the customer service departments of firms across a broad range of industries. In order to select the 170 participants for the survey, Research Now, a third-party research partner of Software Advice, narrowed down a larger group of possible participants to just those who had actually implemented self-service channels in their business and who also had direct knowledge of how the business measured the success of those channels.

It turned out that the most commonly offered customer self-service channels are FAQs and Knowledge bases. Interactive Voice Response (IVR) phone systems comprised the next most common channel. The least commonly offered self-service channel turned out to be virtual agents / virtual assistants. Surprisingly, though, over 50% of those surveyed indicated that their companies had virtual assistants. I would have expected the percentage to be lower, given that virtual assistants are still an emerging technology. Then again, the prescreening narrowed the participants down to those who are already fairly advanced in their use of self-service.

The next major point of inquiry was whether the survey participants monitored the effectiveness of their various self-service channels. They were also asked to provide input on what metrics they used to monitor performance and how effective they considered the metrics to be. About 60% of respondents said that they formally tracked the effectiveness of virtual assistants. I would have expected the number to be closer to 90% or more. To the best of my knowledge, most virtual assistant vendors offer out of the box metrics with their solutions. One type of easily implemented metric is a simple yes/no survey at the end of a chat session that asks the user if the virtual assistant answered his or her question. This user satisfaction metric was indeed the measure that survey respondents employed most frequently. A second metric could be whether the assistant found a response to the question in the knowledge database (or on the company website) or if it came up empty. This type of metric is generally captured as part of the virtual assistant’s conversation log file.

Of the respondents who said they tracked user surveys related to virtual assistant interactions, just shy of 75% said they were satisfied that this is an effective performance gauge. To me, that means there is still room for improving how we measure the reliability of virtual assistants and true customer satisfaction levels. It would be beneficial to have a more accurate, less intrusive method than having to ask the customer if the assistant gave them a useful answer.

A final area explored by the survey was the overall effect of self-service channels on the performance of live customer contact centers. As would be hoped, it turns out that when customers have access to self-service channels, fewer of them call the support desk. As a result, live customer support personnel can take the time to improve the service they give to customers who do call in. By lessening the burden on live support agents, self-service channels helped the majority of survey respondents experience measurable improvements in the following areas:

  • Speed to answer calls
  • Cost per contact
  • First-level resolution rate
  • First-call resolution rate
  • Cost per incident

The Software Advice report is proof that self-service channels are the way to go, right? Well, interestingly, the report references a 2013 Zendesk survey that indicates the majority of consumers would still rather speak to a human than use online self-service channels. It’s important to remind ourselves that we still have a steep hill to climb to convince consumers that calling our support centers should be a last resort. As Millennials and Digital Natives comprise more of the consumer population, this preference for contact with real human support personnel may change. But regardless of how user preferences evolve, our virtual assistant technologies need to continually improve to meet consumer expectations. The only way for us to make sure our technologies are effective is to measure their results, and reports like the one from Software Advice provide insights on how to do just that. 

Dom from Domino’s Stars in His Own TV Spot

Back in August, I wrote about Dom, the intelligent pizza ordering assistant embedded in the Domino’s Pizza app. Dom was one of three winners of the Intelligent Assistants Award presented at Opus Research’s 1st Annual Intelligent Assistants Conference held in mid-September. If you saw my previous blog post about 3 Characteristics of Highly Effective Intelligent Assistants, Dom is a great example of an assistant that performs well in all three categories.

DomEarlier this month, Domino’s released a TV commercial that showcases Dom and his voice-assisted ordering features. If you haven’t seen the ad yet, take a moment to watch it now. What I really like about the commercial is the way it pokes fun at Dom’s single-minded focus. It also hints at how much we’ve come to rely on our speech-enabled intelligent assistants.

You get a strong sense of Dom’s light-hearted personality in the TV spot, a personality that represents the pizza maker’s brand very well. The Dom ad made me realize that intelligent assistants (even specialized, branded assistants) have gone mainstream and they’re part of our world. There’s no turning back. Now if I could only get Dom to help me choose between a Honolulu Hawaiian and a Pacific Veggie pizza. Hmmm.

5-Year Old Chats with ELIZA Chatbot. Fun Ensues.

chatbotsWhat happens when a 5-year old converses with an updated version of the ELIZA chatbot? The resulting conversations may be even more amusing than you imagine.  Kieran Snyder, an IT professional who also happens to have a Ph.D. in linguistics, has been teaching her young daughter River about how computers work. She recently decided to help her daughter understand what it’s like to talk to a chatbot.

Synder published the results of River’s first faltering dialogue with a new version of the ELIZA chatbot that Synder coded up herself. The ensuing conversation is a classic example of both the highs and lows of trying to talk to an “artificial intelligence.” There’s a razor thin edge separating a magical sense of human-to-machine understanding from the total frustration of conversing with a brick wall.

It’s a cute story. And it also shows that conversation is a fine art. Even humans need some time to learn how to practice the art well. Some people, if we’re honest, don’t ever learn to pass the brick wall test. I can tell that River’s going to be a great conversationalist though. I bet she’ll also be the kind of person who gets straight to the point!