Interview with Alexander Weidauer on Conversational Technologies

Botanalytics
Chatbots Magazine
Published in
2 min readDec 1, 2017

--

This month’s guest Alexander Weidauer is Co-founder and CEO of Rasa, which is the leading open source conversational AI company for the enterprise based in Berlin. Enjoy reading!

Alexander Weidauer

Hi Alexander, Let’s start with your research interests. What are your main research topics/main focus while developing conversational software?

Based on what we hear from our enterprise customers and our involvement in the research community, the two big topics are how to make NLU more robust against typos and how to allow for more complex dialogues. The latter is closely connected to out of scope detection, i.e. being able to understand that a user request cannot be handled by the AI.

How promising is current NLP technology according to your views? What parts need to be improved and what other parts are doing perfectly fine?

Bots are not only about “NLP” in the classic sense. Very rarely, conversations are only simple one-shot interactions with one question that just needs to be mapped to the right intent / answer pair. Usually, you have multiple turns and need to take context into consideration. Most of the NLP technologies out there don’t really take this into account. So the ability to converse in more complex ways needs to improve. That’s where our recently launched product Rasa Core comes in.

How do you define best conversational product? What makes a great product from your perspective?

In my view, it is a product where you initiate the interaction with a conversation. For me, it would still count as a conversational product if you see some form of GUI throughout the interaction (e.g. a list of flight options) but it has to start with a conversation and allow the user to text/say anything she wants.

A great conversational product then acknowledges that it cannot answer EVERYTHING — it answers every in scope very fast and delights the user and gracefully handles requests that are out of scope. This can for instance be by handing the conversation to a human agent.

What should be the inner motivation to deploy AI technologies for enterprises?

We’ve seen a lot of “innovation” projects in the last 18 months and people/enterprises building bots for the sake of building a bot. We believe that the motivation should not be to “look innovative” but to actually solve a real business problem. That usually boils down to make more money or save money.

We will be very pleased to have you as our next guest for this enjoyable interview series. If you’d like to participate in our series, shoot an e-mail to hello@botanalytics.co !

And don’t forget to sign up for our newsletter to stay in the loop with the latest news!

--

--