Conversational AI: Data story-telling for a Speech-driven world
Download as PPTX, PDF1 like452 views
The document discusses the significance and implementation of conversational AI (CAI) in data science, highlighting how voice interfaces improve data storytelling through personalized and context-driven interactions. It emphasizes the necessity for customization to enhance user experience while pointing out challenges such as language ambiguity and the need for context. Key takeaways include the importance of narrow use cases, using developer kits, and ensuring security and privacy in future CAI implementations.
Conversational AI: Data story-telling for a Speech-driven world
- 1. 1 B Ganes Kesari Co-founder, Head of Analytics Gramener Conversational AI: Data Story-telling for a Speech-driven World Data Visualization NY| 17th Jul ‘18
- 2. Have you ever had a voice conversation with any Digital Assistant or a Smart voice interface? Have you used a Digital Voice Assistant at work? Do you own Alexa, Google Home or Homepod?
- 3. 3 AGENDA • Why Digital assistants for Data Science? • What is Conversational AI? • How to implement CAI for Business? • How did we do it? • Prototype Demo
- 4. 4https://chatbotsmagazine.com/what-is-a-conversational-ui-and-why-it-matters-de358507b9a2 EVOLUTION OF DIGITAL INTERFACES FOR STORYTELLING
- 6. 6 By 2021, more than 50% of enterprises will be spending more per annum on bots and chatbot creation than traditional mobile app development. Gartner “ Photo by Alex Knight on Unsplash
- 7. 7 WHAT IS A CONVERSATIONAL INTERFACE? A Conversational UI enables interaction with the computer on human terms. “
- 8. Have you interacted with a Digital Voice Assistant or a Customer Service conversational Bot? How many of you were delighted with the Experience and would do it again?
- 9. 9 Meaningful insight calls for depth and tailored usecases No winners for the Amazon Alexa Grand Prize in 2017! WHY IS IT SO DIFFICULT? 1. Language can be very ambiguous 3. Conversations demand Context 2. Insight calls for customization “We saw her duck.”
- 10. 10 By 2020, 40% of bot virtual assistant applications launched in 2018 will have been abandoned. Gartner “ Photo by Edu Lauton on Unsplash
- 11. 11 WHAT ARE THE KEY INGREDIENTS OF A DATA SCIENCE CUI? InsightfulNatural Customize d Contextual DelightfulGraceful
- 12. 12 UNBUNDLING AN INTERACTION Voice Interface Voice to Text Text to Intent Intent to Insight NL Understandin g Insight Identificatio n NL Generation Insight to Text Voice Interface Text to Voice Amazon Lex Apple Siri Bing Speech IBM Watson Google Speech Amazon Lex NS Ling. Tagger MS LUIS Watson NLU Google NL Amazon Polly AV Speech Synth Bing Speech Watson T-to-S Google Speech “ Cortana, Siri, Alexa and Friends: Welcome to the AI Conversation!” by Svetlana Sicular, Gartner Summit Texas, 2018 Usecase specific and needs a Custom Build
- 13. 13 CAI AT GRAMENER: A PILOT IMPLEMENTATION Summary Comparison Trend Distribution Crosstab Listing Change Business Usecase: Entities Covered: Intent Handled: Broadcasting Media Viewership and Performance over the past 6 months Show Channel Network Genre
- 14. 14 DESIGN STORYBOARD & WIREFRAME
- 15. 15 ”Alexa, ask Gramex Assistant, How is Colors doing” Lambda Function Alexa Media Skill (Hosted in Alexa Cloud) Application Data Gramex Platform Audio response to Alexa – “Colors has a rating of 52.35 points and has risen by 2.4% over last week.” Visual SCHEMATIC FLOW Voice Interface Voice to Text Text to Intent Intent to Insight NL Understandin g Insight Identificatio n NL Generation Insight to Text Voice Interface Text to Voice Amazon Lex Custom Build Gramex Gramex Amazon Lex
- 16. 16 Entity 2Entity 1Time PeriodIntent DEMO – SCENARIO 1 ”Alexa, which did ? or ?“better last quarter Star Plus Colors (Comparison)
- 17. 17 Entity 1Time PeriodIntent DEMO – SCENARIO 2 ”Alexa, show me the performance of for the last “ 3 monthsColors Trend
- 18. 18 TAKEAWAYS • Experiment with narrow usecases that have high usage • Get started with Developer Kits and APIs • Target to Deliver insights and Drive actions • Build for multiple input (audio-chat-click) and output (audio-screen) channels • Design for security and privacy (not just convenience) The Future of Conversational AI Things to plan for when you implement CAI: Ubiquitous Invisible Essential
- 19. 19 @kesaritweets @kesari INSIGHTS Extract meaning using automated patterns AI & MACHINE LEARNING SERVICES VISUAL NARRATIVES STORYTELLING Creative ThinkingCritical Reasoning SOFTWARE THROUGH SERVWARE: augmenting human intelligence with technology Binding visuals together into a logical story GRAMENER IS A DATA SCIENCE COMPANY THAT SIMPLIFIES DATA CONSUMPTION THANK YOU!
Editor's Notes
- #8: A paradigm shift from earlier interfaces that call for syntactical commands or clicking icons.
- #10: NLU is an AI-Hard problem Conversation – Context – Dialog management
- #13: Conversation engine: Displaying the right results visually Audio feedback in the form of voice response based on the User Input Computation Understanding the Intents and Entities Querying the right computation based on the Intents and Entities requested NLU Understanding the user input Extracting the user input into Intents and Entities