Broadcast Talking Heads of Your Favorite Celebs

HeadcastLabThere’s a lot of potential for virtual avatars. A recent Techcrunch article described HeadcastLab, the makers of Headcast branded avatars for popular personalities.

The idea behind the Headcast is that avid fans will be interested in getting updates from a lively avatar version of their favorite celebrity. Media stars and other well-known persons can work with HeadcastLab to create cute, cartoon-like images of themselves.

Once these avatars are ready for prime time, the real live star can record pithy tweet-like communications and broadcast them out to followers. Instead of just getting a few dry lines of text, the fan community can watch a pleasing cartoon version of their idol talk to them as if they were skyping with a pal.

The first guinea pig celebrity is Stephen Fry, a British actor, author, and television personality.

A Little Birdie Told Me – NLP Mines Book Recommendations From Twitter

ParakeetsImagine this. You crawl into bed at night and drift off into a deep sleep. While you’re floating through dreamland, your ever wakeful personal virtual agent is sifting through hundreds and hundreds of twitter feeds to find out what books people are talking about. Not only is the agent smart enough to know when a tweet mentions a book title, it can also surmise from the tweet context whether the comment about the book is positive or negative.

You wake up in the morning, shuffle into the kitchen, and find a fresh pot of coffee already made.

“What’s new today?” you ask your personal virtual agent.

“It’s a lovely day,” it responds. “You don’t have any meetings until noon. And I found  three books that I think you’d really be interested in reading. Would you like to see them?”

It might not be long before this scenario becomes reality. Parakweet is a new company that has launched several Twitter information mining products. One of the products is called and it uses natural language processing technology to parse and comprehend tweets about books. The same technology is also used to gather together movie recommendations. The technology challenges that Parakweet has overcome aren’t trivial. Lots and lots of people tweet about books. But how can a NLP agent really detect if what they’re saying about the book is positive or negative? Parakweet CEO Ramesh Haridas says that their technology has tackled this challenge.

Users can sign up for a account and have book recommendations sent to them. It remains to be seen how long it’ll be before the technology is integrated into a coffee-making intelligent personal agent!

An Emoticon For Every Occasion – And Maybe One For Virtual Agents

EmoticonsVirtual agents need to understand and respond accurately to customer questions. That’s a given. But the day’s coming when our virtual agent friends will need to be able to express emotion. In a previous post, I wrote about research into the importance of avatars and how an agent’s physical representation can influence how people interact with it. Research is underway into embodied conversational agents and how to enable them to express human-like emotions.

But what about simple emoticons? We use them so often these days and depend on them to help us convey, or understand, how the friends we interact with are feeling. According to a recent article in Popular Science, Facebook believes enough in the importance of emoticons to invest in emoticon research and development. Working together with UC Berkeley psychology professor Dacher Keltner, Facebook experiments gauge  how emoticons affect a person’s response to specific messages or requests. For example, experiments show that friends are more likely to comply with a request to remove an offending photo or comment when the requester includes an emoticon that gets their feelings across.  Facebook is trying out different expressive representations and adding them to the collection of available Facebook Stickers.

It’s not clear how emoticons might be used in junction with virtual agents. Perhaps a customer-facing agent could learn to include appropriate emoticons when replying to customer questions. Maybe a sad or sympathetic face if the customer says that an item didn’t work for them and they need to return it? Or maybe I should leave that decision to the professionals! Either way, it’s a pretty sure bet that emotional expression will play a role in virtual agent interactions of the future.

Guess Who? – Speaker Verification Technology

Who Am IThis week there was news that Barclays Wealth & Investment Management has implemented a speaker verification solution from Nuance. Barclays is using Nuance’s FreeSpeech product to verify a customer’s identity without asking for a password or digging for answers to personal questions. How does the technology work?

A customer calls into a contact center and starts up a conversation with a human customer support agent. FreeSpeech listens in the background and biometrically analyzes the speaker’s voice pattern. It matches the voice against a stored audio file that was built for that customer during earlier calls. During these subsequent interactions with the call canter, the FreeSpeech voice recognition software can usually verify the customer’s identity within less than half a minute into the conversation. The process is completely transparent.

Detailed documentation describing the Nuance FreeSpeech product indicates that the software works regardless of language or accents. It relies strictly on the individual’s unique voice print. In a case where the system cannot verify the person with a high enough level of confidence, the human customer service agent is notified and alternative methods of identification (such as asking for answers to stored questions) can be used.

Research shows that many people would rather scrub toilets than come up with new passwords they have to remember. No kidding. If biometric technology exists that can minimize the user’s pain of authentication, then it seems like a worthwhile thing to pursue.

Perhaps it’s only a matter of time before customer-facing virtual agents are equipped with speaker verification capabilities, so that they can ensure a web or mobile customer’s identity prior to aiding them with an account transaction. Voice-based authentication might also be a helpful role for enterprise virtual agents that serve employees within a business’s own internal application environment. I’ll be on the lookout for more implementations of the technology.

Deep Learning and Speech Recognition

NEural networkForbes published an interesting interview with Jeff Dean of Google, along with an introduction to the concept of Deep Learning.  Dean’s work using the Deep Learning paradigm has led to a fundamental change in the way Google’s speech recognition systems work. The previous model was an acoustic model, whereas the new approach uses deeply layered neural networks to sequence and categorize phonemes.

Using this new process, Google saw dramatic improvements in speech recognition.

The three keys areas contributing to these advancement are:

  • The Deep Learning architecture
  • The ability to feed systems vast amounts of data for training
  • The availability of extremely powerful computer processing

One of the advantages of the Deep Learning technology is that software can train itself to recognition patterns. This unsupervised learning enables machines to improve their performance without relying on humans to painstakingly point out every identifying trait of the object or sound the machine needs to learn to recognize. Instead of describing to the program every individual feature of a cat, for example, the Deep Learning network is fed with millions of examples of cats. The network teaches itself to recognize the features that typically identify a cat. This unsupervised learning approach turns out to be much more effective in reality. It’s impossible for a human to code every possible feature of every possible cat in every possible position. A Deep Learning network that has trained itself to pick out a cat by looking at millions of examples is less likely to be fooled.

So what does the Deep Learning technology have to do with intelligent virtual agents and web self-service? Though many virtual agents are still text based, the tide is slowly shifting towards voice activated apps. Customers want their questions to be understood. High performing speech recognition technologies may very well be the backbone of customer-facing and enterprise virtual agents of the future.

Nominations for Speech Technology Leaders

AwardsSpeech Technology Magazine is asking for your nominations for the top speech technology companies. Deadline for entries is Friday, May 31, 2013.

The list of award categories provides some insight into the current state of speech technologies and the surrounding marketplace. It’s interesting to see how the various technology and product categories are perceived, and how they’re all undergoing lots of changes in the midst of this extremely dynamic technology environment in which we live.

The 2013 award categories are:

  • Search Engine
  • Self-Service Suite
  • Speech Analytics
  • Professional Services
  • Voice Security
  • Mobile Voice Search

The Mobile Voice Search category description got me thinking. That category is described as one that recognizes “exceptional applications that incorporate speech technologies for search and directory assistance applications on mobile devices.” I’m not sure why Mobile Voice Search was used instead of the more popular Mobile Personal Assistant term. Of more interest, though, is the focus on search and directory assistance. When I think of Siri and Google Now and the host of other speech enabled assistant apps available on iOS and Android today, I don’t know that “search” describes their capabilities or aspirations. “Directory assistance” also doesn’t seem to cover the core features that make these apps so useful.

Speech recognition and search are certainly integral to the performance of mobile personal assistances. All of these apps go beyond search, though, in their attempts to understand what might be important to the end user and to synthesize data to inform the user in ways that go beyond search. A good example of this post-search information synthesis is the ability  of Google Now and other assistant apps to prompt the device owner that she better get going if she wants to be at her next meeting on time, because traffic problems have cropped up. Providing this real-time, heads up type of information goes well beyond search and can’t really be described as directory assistance. But I may be misunderstanding what’s meant by the Mobile Voice Search category. It’ll be fun to see which companies and products receive the award.

If your company has deployed a speech recognition, virtual agent, or any web self-service technology in the past year, you can nominate yourself for an Implementation Award. That would be a great way to recognize the folks involved in the project for all their effort and for the successful outcome.

The Speech Technology Leaders will be announced at the SpeechTek show in New York, which is scheduled for August 19-21, 2013.

Ready to submit your nominee? You can go straight to the Speech Technology Awards 2013 Nomination Form.

The Future of Web Self-Service – A Discussion with David Lloyd of IntelliResponse

Web Self ServiceI recently had an opportunity to speak with David Lloyd, CEO of IntelliResponse, a leading provider of virtual agent technology solutions for the enterprise. One of the topics we discussed was the future of web self-service and virtual agent technologies. David Lloyd has been in the web self-service space for well over a decade. He understands the trends and offered some interesting insights into how the marketplace for intelligent virtual agents might evolve.

While market penetration for virtual agents in a consumer self-service setting is still low, Lloyd sees a very bright future for these solutions. Two key drivers for future growth in this area are consumer expectations and the changing goals of businesses that serve consumers.

Consumer Expectations – Hand Holding Preferred

Not all that long ago in the history of the Internet, the typical consumer expected to have to put some effort into finding things on a business’s website. Websites were often cluttered and clunky. If there was a search box or sitemap, we’d dig in and spend some time figuring out what we needed to do to find the information or product we were looking for. But times have changed. Consumers have many businesses and channels, including web, social, and mobile, to choose from. We no longer expect to have to go hunting when we need information. We want the business to reach out to us and proactively find out what we want and need.

A virtual agent that is knowledgeable about the business and always ready to understand us and answer our questions is a perfect way to meet our evolving expectations as consumers. Presented the choice between a website with a static FAQ buried in the footer and one with a smiling virtual agent eager to answer questions, most of today’s consumers are going to choose the latter.

It’s Not Just About Selling Anymore – It’s About Engaging

We’ve transitioned from the first frenzied wave of social media to an era where social has become intertwined with the fabric of business. Progressive companies that meet their customers through social, web, and mobile channels are looking for new, higher levels of engagement with these customers. The more the business can learn about the people who visit their various storefronts and decide to purchase or not purchase their goods and services, the better positioned the business can be to serve these people effectively. It’s about adding value, and really getting inside your customer’s head enables you to create value a lot better than you could if you were just guessing.

Offering smart virtual agents that can answer a consumer’s questions, instead of leaving that person to fend for her/himself, makes a ton of sense.  Providing a web self-service option through a virtual agent is akin to asking the customer if you can be of assistance, instead of ignoring them. Not only that, you’re learning from the questions that your customers are asking. (More on this great learning opportunity in an upcoming post on the Voice of the Customer). Today’s successful companies want and need to build a relationship with their customers, and offering smart web self-service options can help them reach that goal.

Where Does This Leave Human Customer Support Agents?

During the conversation with David Lloyd, I made a comment to the effect that, given the technological advances in virtual agent capabilities, the future for human call agents was rather dim. Lloyd disagreed. He indicated that in his experience, companies using IntelliResponse virtual agents find that their human customer support agents take on an even more important role.

Since virtual agents do such a great job fielding the many low-complexity questions, human customer support representatives are able to focus more time helping the consumers who have bigger issues. Customer reps are no longer inundated with lots of simple calls, which means the customers who call in with more complex questions don’t have to spend time waiting on the line growing more and more frustrated. Being able to offer this more timely and higher quality support translates into happier customers, fewer negative reviews, and more positive word of mouth for the company.

Based on all of these trends, Lloyd expects the demand for virtual agent solutions and other advanced web self-service technologies to steadily grow. As these technologies become increasingly smarter and more adept at assisting consumers, they may become a necessity to maintain competitive advantage. If your company or organization hasn’t thought about a web-service strategy yet, it’s probably time to start.