How to Build and Design Chatbots (DevDays Vilnius 2017)

Chatbots are rising trend in the tech industry. However, most of them are not really good. This is mainly due to the fact that designing and building them is very hard. But the potential is huge, and the applications immense.

The talk intends to provide advice on how to design and build bots that won’t frustrate the user and will actually build empathy with them.

Get in touch with me via http://danicfilip.com/

Cognito the personal finance assistant: https://cognito.spicefactory.co/

Sava the conference bot: https://savabot.com/

Transcript

1. How to Build and Design Chatbots DevDays Vilnius 2017 @DanicFilip
2. Filip Danić software engineer at SpiceFactory
4. What are Chatbots?
5. What are Chatbots?
6. What are Chatbots?
7. What are Chatbots?
8. Why now?
9. Why now? The home and lock screen are saturated by Messaging apps.
10. Top Platforms?
12. Voice Instead of Text?
13. What can chatbots do now?
14. Big Brands & Chatbots
15. Austrian Airlines
16. Tommy Hilfiger
17. Celebrities?!
18. KFC
19. Western Union
20. My own experience
21. Cognito
22. Sava
23. Categorizing Chatbots
24. Categorizing Chatbots By model Generative Retrieval
25. Categorizing Chatbots By domain Closed Open
26. Open Conversation capability Response mechanism Retrieval Generative Easiest Intelligent bots General AI Categorizing Chatbots Impossible Closed
27. So, let’s design a chatbot
28. So, let’s design a chatbot …for ordering pizza!
29. Requirements
30. Requirements 1. Choose pizza 2. Choose extra toppings 3. Review order
31. Requirements 1. Choose pizza 2. Choose extra toppings 3. Review order
32. Requirements 4. Place order 5. Report delivery time 1. Choose pizza 2. Choose extra toppings 3. Review order
33. Naïve attempt #1
35. Not ideal and not very “smart”
36. Naïve attempt #2
38. Image Description Actions (with defined payloads)
45. Better but looks like an app
46. Comes with all the UX issues of an app and UX issues of using a chatbot Better but looks like an app
47. Moving towards “real” AI
48. Moving towards “real” AI “one large margherita with extra mozzarella cheese and one small pizza mafioza”
49. NLP: Syntax
50. NLP: Syntax
51. NLP: Entities
52. NLP: Intents
53. Contextual Conversations
55. Challenges Moving Forward
56. Training Your Models
57. " <=> # Localization
58. Personality and Writing
59. Security and Privacy
60. Final takeaways