tizen apps with - amazon web services€¢ general preferences (time), short-term interests...

32
1 Tizen apps with Context Awareness, powered by AI by Shashwat Pradhan, CEO Emberify

Upload: hoangnga

Post on 01-Apr-2018

216 views

Category:

Documents


2 download

TRANSCRIPT

1 Tizen apps with Context Awareness, powered by AI by Shashwat Pradhan, CEO Emberify

2

Context refers to information that characterizes a situation, between: •  Apps •  People •  Surrounding environment

•  Making apps smarter and more relevant to every individual user by understanding them

Introduction

3

Contextual apps understand what’s going on in the user’s digital and physical world

Introduction

4

•  Today, an average smartphone has about 10 sensors

•  Contextual data: Current location, time, surrounding brightness, user activity

•  User’s digital world: Apps being used, Facebook API, Instagram API

•  Wearables & IoTs are bringing in many new data points

Introduction

5

•  Alarm based on weather & traffic to work by location sensing

•  Phone goes on vibrate based on proximity to office or a movie theatre

•  Reminders based on travel tickets on email •  Adaptive UI – App theme changes according to

surrounding brightness

Some contextual experiences

6

•  Sense, understand and adapt

•  Get user data from sensors or social networks •  Build algorithms to understand the contextual

data •  Personalize content & provide proactive

recommendations

Contextual Lifecycle

Sense  

Understand  

Adapt  

7

Example of a context aware alarm app

•  Senses the location of the user •  Understands the current weather & traffic

through an API and usual waking up time •  Adapts by letting the user sleep longer or

shorter based on these conditions

Contextual Lifecycle

Sense  

Understand  

Adapt  

8

Tizen 2.4.0 on mobile came with a large set of Context APIs so developers don’t need to access sensors directly •  Activity Recognition (Wearable Also) •  Contextual History •  Contextual Trigger •  Gesture Recognition (Wearable Also)

Context with Tizen

9

•  Activity Recognition – Stationary, walking, travelling, running with accuracy

•  Contextual History – Device Usage Patterns like app usage, peak time and commonly used settings

•  Contextual Trigger – Based on a contextual event the app or notification is triggered

•  Gesture Recognition – Shake, Tilt, Snap, Orientation and other gestures are detected

Context with Tizen

10

•  Apart from the data from the Context API’s, user data can be obtained from Social API’s

•  Calendar and contacts can provide apps with information about the user’s physical world

•  Using Sensor API’s sensors can be directly polled •  For most cases Tizen Context API’s are doing the

sense and understand part for developers

Context with Tizen

11

•  With Artificial Intelligence, context aware apps can be more proactive and intelligent

•  AI algorithms can help make future context predictions

•  Context awareness can make AI applications like Chatbots smarter

Context with AI

12

•  Example- The chatbot needs to recommend the user places when it’s raining

•  Chatbot: Since it is/is going to be raining outside I am recommending you indoor places

•  Gives the Chatbot more human intelligence properties considering the user’s context

Context with AI

13

AI can help with contextual apps with: •  Sensing with abstraction of data •  Sensors can generate huge amounts of data from

which AI algorithms can help extract the relevant data

•  Also, understanding data through audio, images and video can be done through AI branches like computer vision & speech recognition

Context with AI

14

•  User Profiling •  Gathering preferences of a user through sensors,

behavior, social networks or even explicitly •  Personalizing the content and Adaptive nature of the

app according to different user profiles •  AI techniques like association rules or case-based

reasoning can be used

Context with AI

15

•  User Profiling •  Example- News app profiles users to give relevant

articles based on user’s interests •  General preferences (Time), Short-term interests

(Sporting event) & Long-term interests (Politics) •  Keywords can be weighted to prioritize stories

Context with AI

16

•  Context Reasoning •  Based on the context situation an app may need to

adapt •  Examining the contextual information and making a

decision based on rules and logic •  The decision logic can further evolve with Machine

Learning algorithms

Context with AI

17

•  ML algorithms learn from and make predictions on data

•  ML algorithms work on models have to be made based on sample inputs

•  Enables context prediction – which sensor data could be most important in the future

Context with Machine Learning

18

•  Using a combination of sensors, Machine Learning models can be used to determine user activity

•  Extract sensor data and train ML models •  Multiple context data used together can give more

specific information about the user •  Example Accelerometer & Barometer can be

used together to detect walking vs cycling

Context with Machine Learning

Sensor Data  

Machine Learning  

Server / On-device model  

19

•  ML algorithms make sense of noisy/conflicting data from sensors

•  Large datasets are useful to train & fine tune Machine Learning models

•  ML algorithms use raw sensor data to churn out signals based on training models like high level activities

Context with Machine Learning

20

Six technology forces powering contextual apps: •  Mobile •  Social Media •  Sensor evolution •  Cloud & Big Data •  Wearable & other IoTs •  Artificial Intelligence

Technology Powering Contextual Apps

21

•  Launchify – App recommendation widget •  Predicts which app the user needs right now in the

widget

•  Context signals measured for prioritizing information: •  User travelling •  App Usage Patterns •  Location

Case Study

22

•  Launchify – App recommendation widget •  Using Machine Learning algorithms to learn based

on user behavior •  Simple weighing algorithm to give each contextual

parameter weight to priorities •  Senses where, how long, how often, what situations

are the apps being used

Case Study

23

•  Some common sense assumptions are needed in addition to the sensor data based on general human behavior to get more accuracy

•  Sometimes sensors can give us conflicting data. •  Use multiple sensors to confirm it •  Simple logic can be applied to the algorithm like

repeating of a certain event occurrence before counting it to avoid random events

Experiences with Contextual Apps

24

•  Proactive Recommendations •  Recommending the user outdoor places on their lockscreen if

there's a chance of rain

•  Lifelogging •  Quantified Self apps to track the user’s life automatically

•  Adaptive User Experience •  Automatically changing the theme according to sorrounding

brightness

Use Cases

25

•  Context is the secret sauce making an app smart & unique

•  Foursquare doubled down on locations services to give proactive recommendations of food when you’re sitting at a restaurant

•  The contextual fabric can provide a personalized experience

•  New value for users can help apps find interest in App Stores crowded with millions of apps

Use Cases

26

•  Contextual data is not always accurate

•  Allow the user to correct and edit the contextual data

•  Eg. Slow driving is often confused as cycling

•  Machine Learning models take huge amounts of data to train for accuracy

Limitations

27

•  With Context Aware apps you need to be transparent what the app is doing with the user’s data

•  There needs to be a clear privacy policy •  User’s should be able to disable the services •  Encryption and security protocols need to be in

place

Limitations - Privacy

28

•  Rather than providing the wow factor some contextual apps go over the freaky line

•  Nokia’s Trapster (Similar to Waze) would allow it’s users to stalk other users accurately

•  Huge user privacy & trust issues

Limitations - Privacy

29

•  Sensors and background services can consume lots of battery life

•  Data should be polled on triggers rather than a timer

•  Rather than going to the sensor every time it would be more efficient to get data through an app that just polled the data

Limitations – Battery life

30

•  More IoTs and wearables will bring in new sets of data and better quality too

•  Smart Cars & Smart Homes will also add to user information

•  Apps will be more automatic with better sensing •  Contextual Apps will be more proactive in nature •  Smartphone OS’s will take more contextual

information to become more intelligent

Future

31

•  Apps will be ‘Headless’, will require minimum interface interaction

•  Smart Notifications, voice and chatbots will be the new interfaces since apps will need lesser input with contextual information

Future

32

Thank you [email protected]

@shashwatpradhan http://emberify.com

Questions