Building a conversational AI using open source tools

Slides for the talk at Open Data Science Community New Delhi meetup.

Transcript

3. Building a conversational AI with open source tools Alkesh Srivastava, Voiceano @_alkesh_
4. What is a conversational AI?
5. New method of human-computer interaction powered by the most natural form of conversation
6. A brief history
8. 1. Simple 2. Concrete 3. Memorable 4. Emotional
9. Computers are only 40 year old
10. Different types
11. *Blatant copy paste
12. Some use cases
13. 1. Brand re-engagement 2. Retail sector 3. Language learning 4. Customer support
16. Retrieval-based
18. Contextual awareness?
20. The problems?
21. 52% people believe that their personal information or data is not secure - Microsoft Voice Report
22. 41% people worry that their device is actively listening and/or recording them - Microsoft Voice Report
23. Ask the right questions
24. 1. Which platform are your users on? 2. What kind of solution you want? 3. UX or UI? 4. Support for different platforms 5. Escalation protocol?
25. Design principles
26. 1. Pick the right evaluation metric 2. Compare with previous user experiences 3. Be ready to get personal 4. User testing
27. Conducting usability testing?
28. Well, double it!
29. Expect the unexpected
30. This isn’t flying, it’s falling with style! - Buzz Lightyear
31. Break
32. Rasa
33. An open source, highly scalable ML framework to build conversational software ‍
34. Why Rasa?
35. 1. No network overhead 2. Cost reduction 3. Control over data 4. Avoid vendor lock in 5. Open source 6. Hackable and highly extensible
36. Under the hood
38. Rasa NLU
39. Natural language processing tool used for intent classification and entity extraction
40. *Blatant copy paste
41. *Blatant copy paste
42. Rasa Core
43. A dialogue engine which uses ML model trained on sample conversations to decide the next step
44. Demo
45. Thank you Alkesh Srivastava, Voiceano @_alkesh_ alkeshsrivastava.com