5. NLP Tools for Chatbots
LUIS
IBM Watson Assistant
Amazon Lex
6. Working Principles
●
Intents
– Mapping of user input
to actions
– min 10-20 example
utterances per intent
●
Context
●
Slots / Entities
– Required information
for current action
– Locations
– Date/time
7. Where to start?
●
Purpose / use case?
●
Supported languages?
●
Platform? -- limitations and features
●
Data? -- previous (chat) interactions
8. ●
When using external NLP tools, keep your dev-
staging-production-cycle
●
Adding voice support adds extra errors
90%-90%-90% -> 73%
●
Self improvement is tricky to get right
(remember Tay?)
●
Manage expectations, better to start simple!
9. Personal examples
●
Tune: Slack tunebot
https://help.tune.com/marketing-console/slack-tunebot/
Go, Python, NLP by Dialogflow
●
Merck bot
Internal tool for interacting with Salesforce
MS Bot Framework, LUIS
10. Reading Materials
●
Chatbots Magazine: chatbotsmagazine.com
●
Chatbot's Life: chatbotslife.com
●
Recast.AI newsletter
●
The UX of Chatbots: uxofchatbots.com