In this session, we'll cover 10 lessons we had to learn - the hard way - by building a AI-powered assistant: Ozlo.
We've been working on Ozlo for over 3 years now. And, as the AI and conversational search landscape has grown over the last few years, we've evolved our thinking from the days of a blank search box. But, what does it take to build a truly compelling conversational search experience?
In this session, we'll cover a breadth of technical strategies that we consider to be the pillars of AI-powered bots:
- How "big" is "big data"? What does it take to ingest, resolve and reason about the 100s of millions of entities in the world?
- What NLP tools and techniques make the most sense to deploy in an environment where users are talking, not typing?
- What does a dialog system that fulfills the need for a fluid, back and forth, conversation with the user look like?
- How does such a system generate natural sounding, and grammatically correct responses back to the user?
- Finally, how do we measure all this to make sure the user is having a great experience with your product?