SXSW2009 - Collaborative Filters Recommendation Engines
- Anton Kast listing common examples.by thody at 3/14/2009 10:03:07 PM
- Spam filters, pagerank, tagging, comment moderation, help systems, facebook adsby thody at 3/14/2009 10:03:45 PM
- Thumbs up/down on everything.by thody at 3/14/2009 10:04:16 PM
- Recommendation: personalized. Custom search results, Amazon's recursive filtering of recommended productsby thody at 3/14/2009 10:05:13 PM
- Going over digg.com recommendation engineby thody at 3/14/2009 10:06:04 PM
- Comparison of homepage vs. recommended stories. Correlated with users "like you"by thody at 3/14/2009 10:07:00 PM
- Sparsity problem: submissions can grow faster than active diggersby thody at 3/14/2009 10:07:33 PM
- Gray sheep problem: Small group of people very enthusiastic about unpopular viewsby thody at 3/14/2009 10:08:07 PM
- Erik Frey from Last.fm takes overby thody at 3/14/2009 10:08:31 PM
- Talking about discovering new music through last.fmby thody at 3/14/2009 10:09:30 PM
- Relationships between songs and users and users and other users (people recommendations).by thody at 3/14/2009 10:09:56 PM
- Algorithms and social channels.by thody at 3/14/2009 10:10:03 PM
- Different algorithms for different contextsby thody at 3/14/2009 10:10:18 PM
- Lean Forward recommendation: Engaged in site, interested in varietyby thody at 3/14/2009 10:10:34 PM
- Lean Back recommendation: not engaged, ambient, backgroundby thody at 3/14/2009 10:10:57 PM
- Data is the most important ingredientby thody at 3/14/2009 10:11:07 PM
- Scrobbles + Social Tags + Love + Ban + Skipby thody at 3/14/2009 10:11:31 PM
- 25 billion scrobbles on Last.fm so farby thody at 3/14/2009 10:11:45 PM
- Scrobbles are out of context and don't tell the whole story. Unknown artist found similar to Bethoven because his track was also in default sounds folder in win xpby thody at 3/14/2009 10:14:08 PM
- More data points help form context and improve recommendationsby thody at 3/14/2009 10:14:48 PM
- Scott Brave of Baynoteby thody at 3/14/2009 10:15:04 PM
- Recommendation engine as a service on other websitesby thody at 3/14/2009 10:15:27 PM
- Javascript snippet to gain databy thody at 3/14/2009 10:15:36 PM
- Tracks engagementby thody at 3/14/2009 10:15:48 PM
- Affinity Engineby thody at 3/14/2009 10:15:55 PM
- Re-orders search results based on user patternsby thody at 3/14/2009 10:16:36 PM
- Too Many Results: technically correct results, but practically uselessby thody at 3/14/2009 10:17:13 PM
- 50% give up after third resultby thody at 3/14/2009 10:17:29 PM
- Fulltext matches keywords, but doesn't yield useful resultsby thody at 3/14/2009 10:18:41 PM
- Full-text processing, Meta-tagging Taxonomy, Folksonomyby thody at 3/14/2009 10:19:25 PM
- Implicit Tagging: tracking keywords through user actionsby thody at 3/14/2009 10:20:29 PM
- David Mayer Roberts of The Filterby thody at 3/14/2009 10:21:16 PM
- Entertainment content recommendation engineby thody at 3/14/2009 10:21:41 PM
- bayesian collaborative filter modelby thody at 3/14/2009 10:21:55 PM
- UI Examples: Taste Cloud: visualization of implicit and explicit actions, infers other elements, creates feedby thody at 3/14/2009 10:24:12 PM
- Great visualization: data points surrounding dart board like heart. Dearest items are shown closest to the center of the heartby thody at 3/14/2009 10:25:06 PM
- Anonymized data to find relationships between verticals (music, video, etc)by thody at 3/14/2009 10:26:07 PM
- Notion of an entertainment DNAby thody at 3/14/2009 10:26:48 PM
- John Sanders of Netflixby thody at 3/14/2009 10:27:41 PM
- 60% of movie selections based on recommendationby thody at 3/14/2009 10:28:36 PM
- First: The rating widgetby thody at 3/14/2009 10:29:39 PM
- Second: Take score and sort implied interestby thody at 3/14/2009 10:30:27 PM
- Next: Built in k-nearest-neighbour algorithm to tie in movie likenessby thody at 3/14/2009 10:30:52 PM
- Next: Added interest-based discovery + meta data connections (same actor/director)by thody at 3/14/2009 10:31:21 PM
- Added genre ratings to explicitly collect interestsby thody at 3/14/2009 10:31:51 PM
- Found it valuable to explain why something was recommended based on past evidence.by thody at 3/14/2009 10:32:34 PM
- Recommendations are domain specific, and need to be tailored to different forms of mediaby thody at 3/14/2009 10:34:02 PM
- People want to drive and not be lead. Support actions vs. lead actions.by thody at 3/14/2009 10:34:34 PM
- Lots of questions coming...by thody at 3/14/2009 10:36:53 PM
- Popularity bias skews relationshipsby thody at 3/14/2009 10:37:38 PM
- Book recommendation: Collective Intelligence from O'Reillyby thody at 3/14/2009 10:38:22 PM
- Engines are built on similar priciples but are difficult to port due to the importance of contextby thody at 3/14/2009 10:39:14 PM
- Domain specific tweaksby thody at 3/14/2009 10:40:22 PM
- Talking about global activities contributing to recommendations across platforms. ie. can netflix data influence last.fm recommendationsby thody at 3/14/2009 10:42:01 PM
- Question about filtering data which throws off the recommendation engine.by thody at 3/14/2009 10:43:32 PM
- Discussing positive vs negative input: Thumbs up and thumbs down vs positive only input.by thody at 3/14/2009 10:45:58 PM
- Thumbs up UI. Does thumbs up mean support for an item, or for the review?by thody at 3/14/2009 10:46:31 PM
- AB Testing against single measurementby thody at 3/14/2009 10:49:15 PM
- Geo-awareness is a good place to start if you don't know anything about someoneby thody at 3/14/2009 10:54:44 PM
- playground.last.fm has a list of the guiltiest scrobblesby thody at 3/14/2009 10:55:59 PM
- and we're done.by thody at 3/14/2009 10:56:22 PM
- Who's Blogging
-
thody
- Translate
- Options
-
Comments
Turn viewer comments off and on.
Turn comments off
- Share
