Successfully reported this slideshow.
We use your LinkedIn profile and activity data to personalize ads and to show you more relevant ads. You can change your ad preferences anytime.
Conversational Interfaces and
Machine Learning
@
Martin Mitrevski, ICT Innovations, Skopje, 2017
Agenda
• Conversational Interfaces
• Siri use cases
• Natural language understanding with api.ai
• Text recognizer
UI Evolution
Booking a ride
https://vimeo.com/235740121
SiriKit
• Apple’s framework for accessing Siri from third
party apps
• Apple does the natural language understanding
• Res...
Paying Grocery List
https://vimeo.com/235739858
User input
Speech recogniser
Transcribed text
Intents and entities
Persist extracted data
Present to
the user
The process
api.ai
• Google’s NLU platform
• Accessible as a REST Service (JSON)
• Training is done on the web app by the
developer
• ...
Dialog structure
• Intent - mapping a phrase to a specific action
• Entities - parameters of the action (e.g. location,
da...
Text Recognizer
Implementation
• Vision - iOS framework for real time object detection
and tracking
• Core ML - iOS framework for integrat...
Thank you!
• Questions?
martinmitrevski.com
Upcoming SlideShare
Loading in …5
×

Conversational Interfaces and Machine Learning

17 views

Published on

My presentation at the ICT Innovations conference (http://ictinnovations.org) in Skopje, September 2017.

Published in: Software
  • Be the first to comment

  • Be the first to like this

Conversational Interfaces and Machine Learning

  1. 1. Conversational Interfaces and Machine Learning @ Martin Mitrevski, ICT Innovations, Skopje, 2017
  2. 2. Agenda • Conversational Interfaces • Siri use cases • Natural language understanding with api.ai • Text recognizer
  3. 3. UI Evolution
  4. 4. Booking a ride https://vimeo.com/235740121
  5. 5. SiriKit • Apple’s framework for accessing Siri from third party apps • Apple does the natural language understanding • Restricted to pre-defined domains • Not that customisable
  6. 6. Paying Grocery List https://vimeo.com/235739858
  7. 7. User input Speech recogniser Transcribed text Intents and entities Persist extracted data Present to the user The process
  8. 8. api.ai • Google’s NLU platform • Accessible as a REST Service (JSON) • Training is done on the web app by the developer • Supporting agents in different languages, intents, entities, web hooks, contexts
  9. 9. Dialog structure • Intent - mapping a phrase to a specific action • Entities - parameters of the action (e.g. location, date, product, etc.) • Session - the whole conversation • Context - intermediate states and parameters from previous expressions
  10. 10. Text Recognizer
  11. 11. Implementation • Vision - iOS framework for real time object detection and tracking • Core ML - iOS framework for integrating trained machine learning models • Convolutional Neural Networks (86-90% accuracy) • Cropping the image into separate parts for each letter • Padding insertion, resizing, grayscale conversion
  12. 12. Thank you! • Questions? martinmitrevski.com

×