user profiling by social curation by geng xue supervised by prof. chua tat-seng

Download User Profiling by Social Curation By GENG Xue Supervised by Prof. Chua Tat-Seng

If you can't read please download the document

Upload: tyrone-horton

Post on 25-Dec-2015

216 views

Category:

Documents


0 download

TRANSCRIPT

  • Slide 1
  • User Profiling by Social Curation By GENG Xue Supervised by Prof. Chua Tat-Seng
  • Slide 2
  • A Crowd of Social Network Platforms
  • Slide 3
  • Changes in Concerts & Pope Inauguration 19902010
  • Slide 4
  • Exponential Multimedia 2013 Internet Trends http://www.kpcb.com/insights/2013-internet-trends http://www.kpcb.com/insights/2013-internet-trends The Visual nature of the web increases exponentially
  • Slide 5
  • One of Kind Big Multimedia
  • Slide 6
  • How to Deliver Meaningful Contents to the Right Person ? User Profiling
  • Slide 7
  • Definition A process to establish user profiles by extracting & representing the characteristics and preferences of users. Better Service Better Experience
  • Slide 8
  • Recommendation Recommend ? Similar A B Recommend !
  • Slide 9
  • Recommendation Similar Basic info & Social relationships A B
  • Slide 10
  • So, User Profiling by Multimedia Analysis
  • Slide 11
  • However, Multimedia data are very diverse & unorganized. Traditional approaches fail.
  • Slide 12
  • Solutions Structured multimedia & Social Intelligence. Flickr GalleriesFacebook Like
  • Slide 13
  • Social Curation Modern Funnel: Social Curation. People select organize keep track of items they like.
  • Slide 14
  • Social Curation Services (Pinterest)
  • Slide 15
  • Boards to which image are re-pinned Board name and holder
  • Slide 16
  • Why SCSs good? Organized Contents Content-centric network Better Content Models Refine Content Models
  • Slide 17
  • Why SCSs good? Organized Contents Better Content Models
  • Slide 18
  • Why SCSs good? Content-centric network Social intelligence to refine content models
  • Slide 19
  • Framework Profile Structure (Ontology) Learning Profile Structure Refinement User Profiles
  • Slide 20
  • Ontology-based user profiles (e.g., Fashion Domain)
  • Slide 21
  • Profile Ontology Construction
  • Slide 22
  • An example (pegged pants)
  • Slide 23
  • Profile Ontology Learning Idea : sibling samples are more visually similar, classifiers should be more distinct. V dresses, Strapless dresses, Halter dress, One Shoulder dress
  • Slide 24
  • Failure Cases Features may be Wrong
  • Slide 25
  • Refinement of user profiles Organized contents without social intelligence (content-centric network). Social intelligence to refine user profiles.
  • Slide 26
  • If images are shared by more users/boards simultaneously, they more likely belong to the same preference. User/Board-level Connection
  • Slide 27
  • Observation: Shopping Sites Recommendation T shirt People also see these
  • Slide 28
  • Observation: Movie Recommendation Recommendation
  • Slide 29
  • If images are shared by more users/boards simultaneously, they more likely belong to the same preference. User/Board-level Connection
  • Slide 30
  • Content-level Connection Similar images share similar visual cues and semantics. More Similar
  • Slide 31
  • Mathematical Social Intelligence
  • Slide 32
  • Refinement of User Profiles Multi-level connections are incorporated into the low-rank method Before refinement After refinement User-level connection Bundle-level connection Visual-level connection Semantic-level connection
  • Slide 33
  • Visualization of User Profiles It is a vector: (, 0.13, , 0.23, , 0.3, )
  • Slide 34
  • Experimental Data Data collection 1,239 users, 1,538,658 images. Profile learning and refinement We split labeled images equally into training/testing sets as the ground truths. Image recommendation We split the dataset by pin-time for training/testing We added half noisy data out of fashion domain to simulate real world recommendation system.
  • Slide 35
  • Evaluation of Profile Learning Failure Cases: a)Some image samples are too fine-grained. b)Some concepts tend to co-occur in the same image frequently.
  • Slide 36
  • Evaluation of Profile Refinement Failure Cases: a)Sparse & noisy connections from some outdated items. b)Some items are co-repined leading to similar multi-level connections
  • Slide 37
  • Evaluation of Image Recommendation
  • Slide 38
  • Conclusion Social Curation is NEW! It has Well-organized Contents Social Intelligence We test it on Pinterest (fashion domain).
  • Slide 39
  • Thank You