aOS Moscow - E1 - Bots and language understanding - Rick van Rousselt
1 like48 views
This document discusses the basics of language understanding and chatbots. It covers how language begins with shared concepts and references between people. It also discusses how conversations have implicit rules and coordination of meaning. The document then discusses the current state of chatbots, including purposeless mimicry agents, intention-based agents, and conversational agents. It provides some facts about how users interact with bots, including preferences for user interface controls over text or voice, and how search capabilities can benefit bots. The document stresses starting simple with bots and considering business needs over passing conversational tests.
![The way you say it… is equally important
• You know. I don’t. [So don’t ask me.]
• You know I don’t. [You know that I don’t.]](https://image.slidesharecdn.com/aosmoscow-e1-botsandlanguageunderstanding-190715140843/85/aOS-Moscow-E1-Bots-and-language-understanding-Rick-van-Rousselt-36-320.jpg)
aOS Moscow - E1 - Bots and language understanding - Rick van Rousselt
- 1. 1 aOS Moscow 9/7/2019 Bots and Language understanding Rick Van Rousselt Your picture
- 2. ◦ THANKS TO OUR SPONSORS 2 Web Sharesoft Platinum Silver Diamond
- 3. Chatbots have been around for a long time!
- 8. Once in a Decade Opportunity
- 25. Basics
- 26. 1 2 3 Shared Reference Gärdenfors (2014)
- 27. Language begins with the naming of items in the world A concept is shared between persons and a word is used to name that concept
- 28. Language is more then just words
- 29. perception concept action Our brains link shared concepts to personal emotions and experiences • If somebody says “rabbit then we think about linked experiences • Between each other we understand each other because we share the same experiences
- 30. We negotiate about language and meaning while we are having the conversation Coordination of meaning Coordination of experience Instruction Break Break Confirmation Confirmation
- 31. We negotiate about language and meaning while we are having the conversation A fishing rod is a stick with a string and a hook You can catch a fish with a fishing rod Get me some fish Break Break Confirmation Confirmation
- 32. Maxim of Quantity: Say only what is not implied. Yes: “Bring me the block.” No: “Bring me the block by transporting it to my location.” Conversations have their own rules Grice's Maxims
- 33. Conversations have their own rules Maxim of Quality: Say only things that are true. Yes: “I hate carrying blocks.” No: “I love carrying blocks, especially when they are covered in fire ants.” Grice's Maxims
- 34. Conversations have their own rules Maxim of Relevance: Say only things that matter. Yes: “Bring me the block.” No: “Bring me the block and birds sing.” Grice's Maxims
- 35. Conversations have their own rules Maxim of Manner: Speak in a way that can be easily understood. Yes: “Bring me the block.” No: “Use personal physical force to levitate the block and transport it to me.” Grice's Maxims
- 36. The way you say it… is equally important • You know. I don’t. [So don’t ask me.] • You know I don’t. [You know that I don’t.]
- 39. With great power comes great responsibility
- 40. Not: Should I build a bot But What is the outcome I want for users of my product or service?
- 41. What is the business outcome you want to drive? - Customer retention - Customer service - Ex: helping a user until a live person is available
- 42. Is a conversation al chatbot the best medium for your outcome? Vacation booking site? Recipe's site?
- 43. Does your idea try to replace tasks that people love doing? - Ex: Plan a trip
- 44. Is the cost for using your chatbot-powered product or service less than the perceived value? Ex: Shopping list
- 46. Start easy
- 47. The current state of chatbots
- 49. Purposeless mimicry agents - Psychologist from the ’60s - Illusion of conversations - Lot’s of data and deep learning - Nowadays used mainly for automatic translations
- 50. Recurrent neural network encoding the phrase “How are you?” The result is the vector h3.
- 52. Intention-based agents - Present - Language translated to commands - Alexa, Siri, Cortana - Natural Language Processing (NLP)
- 53. Intents Task or action a user wants to perform => verbs Ex: What's the weather like in Moscow?
- 54. Entities Information relevant to the intent => Nouns Ex: What's the weather like in Moscow?
- 56. Conversation Agents - Future - Multi-turn conversations - Dialog manager
- 57. Bot Facts
- 58. Fact #1: Users love buttons When given the choice of rich controls, text and voice, user will choose: UI Controls first Text second Voice third, only if options above aren’t possible (e.g. Driving a car)
- 59. Fact #2: Users never say things the way you expect As you model your NLP, make sure you’re being realistic
- 60. Fact #3: Search can do wonders for bots • QnAMaker.ai: • Super low friction way of creating a Q&A bot • Built in web and document crawler: Parses data and builds the QnA easily • Not great for too many records • Not great for deep level customization of the experience • Azure Search: • Can work with millions of records like a breeze • Works with different document formats and data sources • Takes more effort to prepare/code/fine- tune
- 61. Fact #4: Not all bots need to have conversations • Sometimes buttons do everything you need. • Avoid the engineer’s urge to complicate simple things • Your business requirement is NOT to pass the Turing Test: It is to solve your user’s problem!
- 62. 11.11.2017
- 63. Q&A Maker Demo
- 66. Chatdown • Transcript generator to generate mock transcripts • Helps non-devs to understand the flow • Helps define a “Happy Path”
- 67. Manage bot resources • MsBot: Create a .bot file that links up different services together • LuisGen: Generate Typescript and C# classes from a Luis export • Luis + QnAMaker : manage LUIS.AI and QnA Maker • Dispatch: integrate multiple LUIS Apps and QnA Services
- 70. Brands need to seriously consider: How will customers talk to our brand? And, who are they talking to? Aiden Livingston
- 72. Marketers craft brand narratives, creative writers plot a storyline, and copywriters make micro-copy magical. Conversation designers are a hybrid of all three.
- 73. Rick Van Rousselt CTO Advantive Belgium Office Apps and Services MVP BIWUG Board Member Email : rick.van.rousselt@advantive.be rickvanrousselt https://www.rickvanrousselt.com jobs@advantive.be