Organizing Chaos: The Growth of Collaborative Filters
summary
Argues collaborative filters are now practical because the web pours in dynamic personal signals (from clicks and links to Twitter, GPS and Nike+iPod) enabling personalization beyond global rank; cites ~7.3 million pages/day in 2003 vs a terabyte/day now, and Dig handling 16,000 submissions a day, planning to ingest ~90 million Facebook users of profile data. Describes using PageRank, clickstreams and cross-site profiles to build recommendation engines that target ads and front pages, and claims this shifts value to personalized relevance.
AI-written summary · about
chapters (6)
- 0:00Introduction to filters
- 1:19Evolution of user data
- 2:18Dynamic web and privacy
- 4:27Filtering approaches
- 5:31Future of personalization
- 7:25Data-driven recommendations
related talks
19mPopping the Filter Bubble
44mHacking the Algorithm: Survive the Era of Digital Cult Worship
20mDesigning for Big Data
05mCameron Marlow at Social Graph Foo Camp, February 2008
28mCrafting Hyper-Relevant User Experiences
19mBig data in the service of reliable news
35mEnhancing User Engagement Through Personalized Recommendation at Dow Jones
39mMachine Learning for Relevance and Serendipity