Tweet
Tweet

Slide 1

Slide 1 text

© 2018, Amazon Web Services, Inc. or its affiliates. All rights reserved. Tailoring Video Streaming Service with Data Jing-Kai Lou, PhD Data Scientist, KKStream, KKBOX Group June, 2018

Slide 2

Slide 2 text

© 2018, Amazon Web Services, Inc. or its affiliates. All rights reserved. KKStream initiated as KKBOX Video Department, spin off in Mar 2016 au Video Pass (Japan), KKTV (Taiwan), giloo (Taiwan), and more coming...

Slide 3

Slide 3 text

© 2018, Amazon Web Services, Inc. or its affiliates. All rights reserved. Confront Many Challenges How to organize content personally for each individual?

Slide 4

Slide 4 text

© 2018, Amazon Web Services, Inc. or its affiliates. All rights reserved. Turn Challenges Into Opportunities People who watch “Midnight Diner (深夜食堂)” also watch? Solutions GET APIs Data Service: finding similar items PUT APIs

Slide 5

Slide 5 text

© 2018, Amazon Web Services, Inc. or its affiliates. All rights reserved. What is Video Streaming Service?

Slide 6

Slide 6 text

© 2018, Amazon Web Services, Inc. or its affiliates. All rights reserved. Pick (content) Pack (encoding) Deliver (CDN) Engage (storefront) Consume (video player)

Slide 7

Slide 7 text

© 2018, Amazon Web Services, Inc. or its affiliates. All rights reserved. What Data Can Support Us? Pick (content) Pack (encoding) Deliver (CDN) Engage (storefront) Consume (video player) social listening, ... perceptual video encoding optimization, ... CDN routing, ... video recommender system, ... adaptive streaming, ... powered with data

Slide 8

Slide 8 text

© 2018, Amazon Web Services, Inc. or its affiliates. All rights reserved. What We Solved?

Slide 9

Slide 9 text

© 2018, Amazon Web Services, Inc. or its affiliates. All rights reserved. Ongoing Journey We develop ML models and features along with operating VOD service. Democratization and Gamification “More like this” on KKTV Intelligent Notification “Discover more” on KKTV Content understanding

Slide 10

Slide 10 text

© 2018, Amazon Web Services, Inc. or its affiliates. All rights reserved. Data Democratization and Gamification

Slide 11

Slide 11 text

© 2018, Amazon Web Services, Inc. or its affiliates. All rights reserved. Intelligent Notification Time prediction AUC 0.88 Engagement +2% Notify according to watching habits

Slide 12

Slide 12 text

© 2018, Amazon Web Services, Inc. or its affiliates. All rights reserved. More like this CTR +13% 深夜食堂 seed more like this Organize content using behaviors

Slide 13

Slide 13 text

© 2018, Amazon Web Services, Inc. or its affiliates. All rights reserved. First bite +7% Discover more Offer feedbacks after interactions

Slide 14

Slide 14 text

© 2018, Amazon Web Services, Inc. or its affiliates. All rights reserved. Rich Display Motivate Users to Act Avoid blur images Focus on faces (sharpness and position) Automatic Extraction

Slide 15

Slide 15 text

© 2018, Amazon Web Services, Inc. or its affiliates. All rights reserved. 66% win ratio lessons learned ● protagonists are preferred ● posture matters Blind Test Result over 150 subjects, 50 episodes

Slide 16

Slide 16 text

© 2018, Amazon Web Services, Inc. or its affiliates. All rights reserved.

Slide 17

Slide 17 text

© 2018, Amazon Web Services, Inc. or its affiliates. All rights reserved. How We Run Projects

Slide 18

Slide 18 text

© 2018, Amazon Web Services, Inc. or its affiliates. All rights reserved. Roles Make Data Projects Run ● Project owner
 represent the business interest; champions of product ● Client
 represent end-user interest; domain expert ● Data scientist
 set and execute analysis strategy; communicate with owner and client ● Data architect
 manage data and data storage; sometimes manage data collection ● Operations
 manage infrastructure; deploy final project result

Slide 19

Slide 19 text

© 2018, Amazon Web Services, Inc. or its affiliates. All rights reserved. Data Project Steps 1. Communication / Request 2. Understand / Hypothesis 3. Data Collection 4. Data Exploration 5. Transformation 6. Model 7. Validation and Evaluation 8. Deployment 9. Operation ● Project owner, Client ● Data architect ● Data scientist ● Operations

Slide 20

Slide 20 text

© 2018, Amazon Web Services, Inc. or its affiliates. All rights reserved. Well-Organized Data Store is Desired

Slide 21

Slide 21 text

© 2018, Amazon Web Services, Inc. or its affiliates. All rights reserved. Our Data Flow Ingestion Modeling Transform Deployment

Slide 22

Slide 22 text

© 2018, Amazon Web Services, Inc. or its affiliates. All rights reserved. Takeaway 挑戰同時是機會，我們⾯面臨臨的恰好正是其他⼈人想解決的問題。 SageMaker 的出現，表⽰示⼀一股趨勢：「⼀一群資料科學家正苦思 如何加速落落實點⼦子」。

Slide 23

Slide 23 text

© 2018, Amazon Web Services, Inc. or its affiliates. All rights reserved. Thank You! [email protected] Find “KKStream” on