'

Nuance expands Cognitive Innovation Group: A look at voice user interfaces, AI, analytics

ZDNet caught up with Tom Hebner, leader of Nuance's CIG team, to talk AI, voice user interfaces, and analytics. Here's a recap of the key themes.

special feature

How to Implement AI and Machine Learning

The next wave of IT innovation will be powered by artificial intelligence and machine learning. We look at the ways companies can take advantage of it and how to get started.

Read More

Video: Why AI is now table stakes in enterprise software

Nuance said it will expand its Cognitive Innovation Group, which is designed to use machine learning and artificial intelligence for customer service, to enable more enterprises to experiment with conversational platforms and analytics.

The company's Cognitive Innovation Group (CIG) initially launched Nuance's Nina platform in 2012 and later rolled out Nina Coach as a human-assisted virtual assistant.

Under CIG's new mission, the group will provide research and know-how for AI and predictive analytics via AI Engagement Services and AI Research teams. Thomas Hebner, who led Nuance's voice user interface and services group, will lead CIG.

CIG will include the following:

  • An AI Lab to match research with needs and use cases for enterprises.
  • Engagement services to create and prioritize roadmaps and advance AI implementations.
  • Research via Nuance's 300-person research team.

I caught up with Hebner to talk AI, voice user interfaces ,and analytics. Here's a recap of the key themes.

The purpose of CIG: Hebner said that the market demand for AI and natural language processing has exploded. Nuance typically has worked with a handful of customers who wanted to conduct research and push the envelope. One of these customers was USAA, which helped hatched Nina. CIG's expansion will allow Nuance to bring more technology from the lab to production, he said. "CIG's AI Engagement Services is for enterprises that want to push the envelope, but don't have the culture where they can take huge risks. It's a more practical application," said Hebner. CIG key customers include USAA, Fidelity, Domino's, and ING Bank.

Must read: Nvidia, Nuance team up to bring AI, machine learning to radiologists | AI and jobs: Where humans are better than algorithms, and vice versa

How will CIG's new services be measured: Hebner said his unit will "plan on driving revenue to the services team," but the returns are more anecdotal. "We want to see what's next and move from a vendor relationship to thought leadership. There's also a potential for new products coming out," Hebner said. "We'll be measured on how many showcase experiences did we deploy in 2018. We're hoping for 4 to 5, but it may not show up on a P&L."

Working with multiple AI platforms: Hebner said CIG's expansion into AI will mean that enterprises may have multiple platforms in play. "Most of the innovative customers have done a Watson pilot," Hebner said. "We're starting with asking what problem you're trying to solve using natural language processing," said Hebner, who CIG will try out third party algorithms, open source and cloud platforms to see what works best. However, Hebner also noted that some industries such as financial services doesn't want data to leave the building.

Must read: Alexa for Business likely to win in smart office| Alexa for Business likely to win in smart office | Alexa for Business: 10 key takeaways | IBM's Watson Data Platform aims to become data science operating system

TechRepublic: Nuance Nina wants to make Alexa more business friendly with its 'human-like' AI assistant | Cheat sheet: Alexa for Business | Amazon AI: The smart person's guide | How to become an Alexa developer: The smart person's guide

How does analytics fit into conversational systems?: "We've built a lot of conversational systems and they revolve around how can I help, who are you and can I solve your problem," explained Hebner. "We want to see how much can we predict about a conversation and get to the end so we can say here is what you need."

Prediction and automated conversations can greatly improve customer satisfaction and get you to the right human, he said. The returns would be minimizing call center misrouting, which adds to enterprise costs. CIG is now training different algorithms to do the prediction and seeing what model works best, he said.

Why your data house has to be in order: Hebner noted that enterprises can be interested in AI, but lack the data hygiene to make it work. "A customer can't do analytics if it doesn't have data. And if you don't have the data you can't use AI," he said. "We will work with them, but it can be an 8 month project to get data organized for analytics."

The great data science hope: Machine learning can cure your terrible data hygiene | TechRepublic: Cheat sheet: How to become a data scientist | Job description: Data scientist (Tech Pro Research)

Voice as the new user interface: Hebner agreed that voice is the new user interface, but also added that conversational UI is hard. Conversational UI falls into the following buckets: Question, answer and dialogue. Here are the key things to know.

  • Question technology is largely mature.
  • Answer technology falls into simple and complex and dynamic. An example of a simple answer technology can go through documentation and provide an answer. "With dynamic answering people think the technology is farther along than it actually is," Hebner said. "To be dynamic you have to have data for complex answers."
  • Dialogue via a voice UI is largely manual. "Someone had to design those voice UI (VUI) flows. Typing up answers takes man hours to create compelling engagement. You have to sit down with business and build a conversation by understand processes, data and build the whole thing. It's a ton of work," Hebner said. "Google, Amazon and Facebook have cool answering technology, but it isn't dialogue.

As a result, CIG is seeing what it can do with machine learning to automate and ingest conversations, said Hebner. "Neural networks can find patterns and ingest conversations," Hebner said. "But at a basic level there's a lot that has to be programmed. We're focused on shrinking the effort it takes to auto learn the dialogue. It will be years until all of that's solved."

Must read: Free PDF download: IT Jobs in 2020, a Leader's Guide

VUI as a acronym: Hebner said that it used the term VUI for years, but it was so abstract that the company dropped it to just identify with UI and UX. Then VUI became cool. "We killed VUI (as a term) and now everyone has a VUI team," he said.

Voice biometrics, the next frontier: At some point in the future, Hebner said a customer will be able to call and a company will know it's you via biometrics and a voice signature as soon as you say hello. Nuance has production deployments of voice biometrics where you can say "my voice is my password." The next frontier will include a conversational biometric approach where you can say, "Hi, this is Larry" and the brand will know how to engage with you. In the labs, Nuance's CIG group is working with passive listening technology that'll run in the background and know your speech enough to give you a quality ID score. "We're not ready for a seamless skip over where your voice is the password (automatically), but we have an algorithm and a secondary algorithm watching the first one," Hebner said. "It's real in the labs, but not ready for prime time."

Must read: Voice biometrics to be the 'front door' for all ANZ bank contact | ANZ bank rolls out Voice ID for mobile banking

Related stories