11. 11
Goal
Video
A
Video
B
Select videos for each video
Video Video Video
Video Video Video
Recommend
Recommend
: :
12. 12
Recommendation by users’ behavior
Select videos watched by common users
TVNW: SBS
Genre: Romance
Year: 2013
TVNW: SBS
Genre: Action
Year: 2014
TVNW: TvN
Genre: Romance
Year: 2013
User 1
User 2
User 3
Watch Watch
Watch Watch
Watch
video
A
video
B
video
C
13. 13
Recommendation by users’ behavior
Select videos watched by common users
TVNW: SBS
Genre: Romance
Year: 2013
TVNW: SBS
Genre: Action
Year: 2014
TVNW: TvN
Genre: Romance
Year: 2013
User 1
User 2
User 3
Watch Watch
Watch Watch
Watch
video
A
video
B
video
C
14. 14
Recommendation by users’ behavior
Recommend “B” on page “A”
TVNW: SBS
Genre: Romance
Year: 2013
TVNW: SBS
Genre: Action
Year: 2014
video
A
video
B
Recommend
16. 16
Issue
Ratio of videos having results : 42%
Popular
Videos
Video
Minor
Videos No results
42%
58%
100%
80%
60%
40%
20%
0%
17. 17
User behavior
Content
attributes
+
Episode / Parts
Other attributes
18. 18
User behavior
Content
attributes
+
Episode / Parts
Other attributes
19. 130,976 videos 79,601 videos
19
Parts
Merge parts into one video
Jungle
Emperor
Leo
Part1
Jungle
Emperor
Leo
Part2
Jungle
Emperor
Leo
20. 79,601 videos 22,844 videos
20
Episodes
Merge episodes into one video
Doctor X
Episode 1
Doctor X
Episode 2
Doctor X
21. 75%
21
Issue
Ratio of videos having results : 42%
Popular
Videos
Video Video
Minor
Videos No results
75%
25%
100%
80%
60%
40%
20%
0%
22. 22
Issue
Remaining 25% How should we do ?
Popular
Videos
Video Video
Minor
Videos No results
75%
25%
100%
80%
60%
40%
20%
0%
23. 23
User behavior
Content
attributes
+
Episode / Parts
Other attributes
24. 24
Procedure
Which attributes are same ?
TVNW: SBS
Genre: Romance
Year: 2013
TVNW: SBS
Genre: Action
Year: 2014
TVNW: TvN
Genre: Romance
Year: 2013
User 1
User 2
User 3
Watch Watch
Watch Watch
Watch
video
A
video
B
video
C
25. 25
Probability of videos having same attributes
TV NW > Country > Genre > Actor, ..
TVNW: SBS
Genre: Romance
Year: 2013
TVNW: SBS
Genre: Action
Year: 2014
TVNW: TvN
Genre: Romance
Year: 2013
User 1
User 2
User 3
Watch Watch
Watch Watch
Watch
video
A
video
B
video
C
26. 26
Application
Select videos focusing on TV NW
TVNW: SBS
Genre: Romance
Year: 2013
TVNW: SBS
Genre: Action
Year: 2014
video
A
video
B
video
D
TVNW: SBS
Genre: Romance
Year: 2014
Rec.
Minor videos
(25%)
27. 27
Application
Select videos focusing on TV NW
TVNW: SBS
Genre: Romance
Year: 2013
TVNW: SBS
Genre: Action
Year: 2014
video
A
video
B
video
D
TVNW: SBS
Genre: Romance
Year: 2014
Rec.
Same TVNW’s videos
Minor videos
(25%)
28. 28
0.13
0.12
0.11
0.1
0.09
0.08
Old New
Click Rate [%]
AB test
Click Rate : +32.4%
It uses AB test frame work ‘Turing’ developed by Ishan
29. It rolled out across the world
29
New recommender
DC
DC
DC DC
Front-end is developed by Huy & Yan Han
33. 33
Contents-based recommendation #1
Fix weights of attribute by user-behavior rec. result
Similarity(video A, video B) =
w1 * Genre Similarity (genre of video A, genre of video B)
w2 * Country Similarity (country of video A, country of video B)
w3 * Actor Similarity (actors in video A, actors in video B)
:
• If attributes are matched, it’s 1.
• Otherwise, it’s 0.
Weights are fixed by user-behavior rec. result
34. 34
Contents-based recommendation #2
• Fix weigh by using user behavior recommeder result
• Estimate similarity for videos which have no results
1.Training 2.Test
Video
A
Video
B
Genre
A
Genre
B
Jaccard
similarity
1v 2v kpop rock 99.0
1v 3v kpop jazz 3.1
1v 4v kpop classic 2.1
Video
A
Video
B
Genre
A
Genre
B
Jaccard
similarity
5v 2v jpop rock ?
5v 3v jpop jazz ?
5v 4v kpop classic ?
Fix weight
Similarity (videoA, videoB ; W)
Estimate jaccard similarity
Similarity (videoA, videoB ; W)