1. Personalized Celebrity Video Search Based on Cross-space Mining
Zhengyu Deng, Jitao Sang, Changsheng Xu
1 Institute of Automation, Chinese Academy of Sciences
2 Chinese-Singapore Institute of Digital Media

2. Outline
Motivation
Framework
Approach
Experiment
Conclusions

3. Motivation
Celebrities are often popular in multiple fields and user interests are diverse.
User 1
User 2
User 3
Sports video
Entertainment video
Interview video
Beckham
like

4. Motivation
Celebrities are often popular in multiple fields and user interests are diverse.
User
Sports video
Entertainment video
Music video
like
Beckham
Bieber
Lady Gaga

5. Non-personalized search
Motivation
Non-personalized search
David Beckham
Daily life
Sports
Interview

6. Problem and solution Motivation Problem Solution
Users have different interest distribution and celebrities have different popularity distribution. How to match user interest and celebrity popularity?
Problem
Learn interest space of users and popularity space of celebrities, then correlate the two spaces.
Solution

7. Framework Query User Interest space Map Associated tags Search engine
Celebrity
Interest space
Popularity space
Map
Re-rank
Search engine
Topic modeling
Associated tags
Query
2019/2/19

8. Approach … … … … … U1 C1 U2 C2 Um Cn Random walk LDA LDA KL-Divergence
User
Interest Space
Celebrity
… … … … …
Vocabulary
Popularity Space U1 C1 Z1 W1 X1 U2 Z2 X2 C2 W2 Um Zp Wx Xq Cn
Random walk
LDA
LDA
KL-Divergence
P(Zi|Ui)
P(Wi|Zi)
P(Wi|Ti)
P(Ti|Ci)
P(Zi|Xi)

9. Approach
Random walk
Vj is the initial probabilistic score; pij is the transition matrix; rk(j) donote the relvence score of node j at iteration k
(1)
Rewrite as
(2)
The unique solution
(3)

10. KL-Divergence Approach
𝑇𝑜𝑝𝑖𝑐 𝑧 (𝑥) is from interest (popularity) space. The KL-Divergence between them is
𝐷 𝐾𝐿 (𝑧 || 𝑥)= 1 2 ( 𝑖 𝑧(𝑖)𝑙𝑛 𝑧 𝑖 𝑥 𝑖 + 𝑖 𝑥(𝑖)𝑙𝑛 𝑥(𝑖) 𝑧(𝑖) ) (4)
where 𝑧(𝑖) (𝑥 𝑖 ) denote the distribution score of topic 𝑧 𝑥 on word 𝑖.
The similarity 𝑠𝑧𝑥 of topic 𝑧 and 𝑥 is defined as the inverse of KL-Divergence.
𝑠𝑧𝑥=1/𝐷 𝐾𝐿 (𝑧 || 𝑥) (5)

11. Video Projection Approach
Given a celebrity video 𝑣 𝑀×1 , project it to interest space Φ 𝐾×𝑀
𝑣 𝐾×1 ′ = Φ 𝐾×𝑀 𝑣 𝑀× (6)
where K is the topic number of interest space. M is the dimension of the vocabulary.

12. Video re-ranking Approach 𝑝 𝑠𝑐𝑜𝑟𝑒 𝑣,𝑢,𝑐) (7)
Given a user 𝑢 and celebrity 𝑐, the score of 𝑣 is
𝑝 𝑠𝑐𝑜𝑟𝑒 𝑣,𝑢,𝑐) (7)
= 𝑖=1 𝐾 𝑃 𝑧 𝑖 𝑣 𝑝 𝑧 𝑖 𝑢 𝑝 𝑧 𝑖 𝑐
= 𝑖=1 𝐾 𝑃 𝑧 𝑖 𝑣 𝑝 𝑧 𝑖 𝑢 𝑗=1 𝐿 𝑃 𝑥 𝑗 𝑐 𝑝 𝑧 𝑖 𝑥 𝑗
where K(L) is the topic number of interest (popularity) space, 𝑧 𝑖 ( 𝑥 𝑗 ) is the 𝑖 th (𝑗 th) topic of interest (popularity) space, 𝑝 𝑧 𝑖 𝑥 𝑗 is approximated by the inverse of KL-Divergence.

13. Data Preparation Experiments Celebrity list
The World's Most Powerful 100 Celebrities List
The 30 Most Generous Celebrities
Top 200 Sexiest Actor
For each celebrity, 200 videos are downloaded from YouTube.

14. User and Celebrity Profiling
Experiments
User and Celebrity Profiling
User  registration info., favorite and uploaded videos  raw tags  stop words  WorldNet  noun tags.
Celebrity  Wikipedia Entry  WorldNet  noun tags
celebrity
user
total
Size
286
200
486
Tags Number
11424
5833
12073

15. Experimental Setting Experiments Experiment data Experiment setup
143 users
106 celebrities
Experiment setup
Each user have some videos related with a specific celebrity. Leave this videos out and learn topics.
Rank this celebrity’s videos for the user.
Evaluation
f-Measure

16. Experiments
Topic simples

17. Doc-Topics distribution
Experiments
Doc-Topics distribution
E.g. Celebrity ”Beckham”
Topic Probability of appearance

18. Topic-terms distribution
Experiments
Topic-terms distribution
E.g. <topic id=“7">
<word weight=" " count="478">jay</word>
<word weight=" " count="457">messi</word>
<word weight=" " count="447">real</word>
<word weight=" " count="438">ronaldo</word>
<word weight=" " count="434">kanye</word>
<word weight=" " count="414">west</word>
<word weight=" " count="413">wayne</word>
<word weight=" " count="396">lil</word>
<word weight=" " count="355">hop</word>
<word weight=" " count="350">lionel</word>
<word weight=" " count="347">beckham</word>
<word weight=" " count="326">beyonce</word>
<word weight=" " count="319">cristiano</word>
<word weight=" " count="316">soccer</word>
<word weight=" " count="316">football</word>
… …

19. Topic-terms distribution
Experiments
Topic-terms distribution
E.g. <topic id=“4">
<word weight=" " count="1382">show</word>
<word weight=" " count="777">david</word>
<word weight=" " count="753">ellen</word>
<word weight=" " count="746">tv</word>
<word weight=" " count="627">comedy</word>
<word weight=" " count="580">jennifer</word>
<word weight=" " count="566">interview</word>
<word weight=" " count="550">degeneres</word>
<word weight=" " count="530">funny</word>
<word weight=" " count="482">letterman</word>
<word weight=" " count="453">hollywood</word>
<word weight=" " count="443">late</word>
<word weight=" " count="416">talk</word>
<word weight=" " count="397">celebrity</word>
<word weight=" " count="362">television</word>
… …

20. Experiments
Different approaches

21. Experiments
Impact of random walk

22. Conclusions Conclusions Future work
We presented a cross-space mining method to exploit the correlation between user preferences and celebrity popularities.
Future work
Instead of returning a ranking list, we will try to visualize the search results into semantically consistent groups.
Investigate the issue of personalized query understanding in more general personalized search applications.

23. Thank you!
Q&A?