Open Source NLU – Alan Nichol, Rasa – Voice Tech Podcast ep.073

Alan Nichol Rasa

Episode description

Aside from having an excellent product, one of the best ways to foster adoption is by building community support early on. One company that has reaped the rewards of taking this approach, is Rasa, an open-source, natural language understanding framework that helps companies build mission-critical contextual assistants, and its co-founder and CTO, Alan Nichol, joins us today.

We kick this episode off by learning more about Rasa’s mission to become the no-brainer technology for building conversational AI. They are achieving this through their relentless commitment to community, developer empowerment, and by what Alan calls a ‘no fairy-dust attitude.’ Building conversational AI is not easy, which is why Rasa is inviting developers along for the journey. From there, we move onto why Alan believes it’s important to get products in front of users at soon as you can. Rather than waiting for full development, if you get what you can into people’s hands and let them use it, the feedback will be much richer, culminating in a better outcome.

We then move into discussing the benefits of open-source. There are almost an infinite number of advantages according to Alan, from increased innovation surface area to the ability to squeeze out extra performance. However, the financial realities are important, and we discuss how Rasa’s enterprise products fulfill these needs.

We round the show off by diving into the five levels of conversational AI and where we currently are, why intents are hindering the scalability of these products, and what’s in the pipeline for Rasa.

Highlights

  • Rasa’s ‘elevator pitch,’ what they do, who they serve, and what they’re trying to build.
  • What Alan learned from reaching out to the community so early on when building Rasa.
  • Learn more about two of the main pieces of Rasa: NLU and entity extraction.
  • Find out more about the training process and how the customer provides the data.
  • Some of the problems that Alan sees with the timeline of most conversational AI projects.
  • Why Alan believes in getting a product in front of people as soon as possible.
  • How Rasa deals with bias, and what they do to improve the datasets.
  • Rasa’s compatibility with larger platforms and why it’s a tenuous space.
  • The main benefits of being open-source, for the user, the developer, and the company.
  • Rasa’s enterprise product and how they view it separate from the technical side.
  • Case studies: Find out about Eddy Travels and McDonald’s use of Rasa.
  • Find out how Rasa has adapted for voice and the work that still needs to be done there.
  • Getting started: What developers can do to dip their toes in the Rasa waters.
  • User management: Some of the challenges that come with managing multiple users.
  • How Rasa supports collaboration and the tools and practices they have in place.
  • The five levels of conversational AI: What they are and where we currently are.
  • Why Alan believes that getting rid of intents is necessary to move to the next level.
  • Alan’s advice for people wanting to enter the AI space, and what’s in store for Rasa.

Quotes from the show

[14:43] If you’re building conversational AI, you use Rasa. That’s where we’re headed.

[13:23] To build a good assistant, build a bad assistant, then give it to people to test.

[24:32] If you think that defensibility is having a secret algorithm, you’re living in a fantasy

[42:06] We’re trying to build a good foundation and then build the right tools on top of that.

Links from the show

About the author

Carl Robinson
Carl Robinson

Carl is the host of the Voice Tech Podcast. Since launching in April 2018, Carl has conducted scores of in-depth interviews with voice industry experts, building one of the most well known media brands in the voice AI space.

He also publishes Voice Chops Tuesday, the number 1 voice technology newsletter, enjoyed by thousands of voice tech fans each week.

Carl is a startup founder, product manager and data scientist, and recently presented a model for voice emotion conversion at ICASSP 2019.

To contact Carl, use the links in this box to send an email, tweet or message.

Share this article

If you found value in this article, please consider helping others by sharing to your network. Just click one of the links below.

What do you think?

0 replies

Leave a Reply

Want to join the discussion?
Feel free to contribute!

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.

Related Posts

Christian Petroske ShoutWorks
Voicetech Alex Miller
Simon Landry Voice Market Data

Get notified about new articles

Upcoming Events


Featured Products