Personally Relevant Content

1

• Personally Relevant Content
• Based On Social Connections

2
Team

• M. Deniz OKTAR, BSc _at_ Koc University, Istanbul
• Firat GELBAL, MSc _at_ TU/e, Eindhoven
• Baris Can DAYLIK, BSc _at_ Koc University, Istanbul
• Selcuk ATLI, MSc _at_ RPI, New York
• Board of Advisors
• Prof. James Hendler, Professor in Computer
  Science _at_ RPI, TWC
• Dr. Tarcan Kumkale, Assist. Prof. in Psychology _at_
  Koc
• Dr. Oznur Ozkasap, Assist. Prof. in Computer
  Science _at_ Koc

3
Problems

• No psychological or social aspect
• Rapidly emerging online media content like news
  feeds need special engine
• Computation time
• New user issue
• Critical mass problem
• Overspecialization
• Explicit feedback

4
Value Proposition

• Wisdom of the Crowd vs. Circle of Trust
• Social added value
• 3-layered Hybrid Engine
• Weighted social proximity graphs
• Multiple social graphs for different interests
• Implicit/explicit feedback

vs
5
Individual Interests
Politics

Sports







Music




Technology


Relevance a Content ß CF ? Social
Relevance a Content ß CF ? Social
6
Network Interests
















Relevance a Content ß CF ? Social
Relevance a Content ß CF ? Social
7
Circle of Trust Interests















Relevance a Content ß CF ? Social
Relevance a Content ß CF ? Social
8
Interests Reconfiguration












Relevance a Content ß CF ? Social
9
Interests Reconfiguration












Sports Proximity
10
Interests Reconfiguration












Politics Proximity
11
Smarter Social Ads

• Social Advertising
• If your friends have it, then you will have it
• People are influenced by their friends
• iletken captures people that most influence their
  friends
• Opinion leaders are easy to find

12
Thanks

• Focus on developing the recommendation system.
• Add value to your social network.
• For more information www.iletken-project.com

13
Questions?

• New User/Critical Mass Poblem
• Social Capital
• Importance of Trust
• Why Different
• Chinese Song Problem
• Social Noise
• Thanks

14
Critical Mass Problem
BACK
Solution Hybrid Engine
15
Trust is Value
BACK
16
Social Capital
BACK
17
Why Different
BACK

• Social Added Value
• Live Social Graphs

vs
live updates on social graphs
18
Chinese Song Problem
BACK
Time
19
Social Noise
BACK

• Iletken uses Social Noise instead of Random

Noise Social Capital
Noise Random(t)
vs