SlideShare a Scribd company logo
1 of 21
Download to read offline
In Other News
Broaden your horizons by
reading like a writer
Charlotte Greenan
Using a social network of
journalists to recommend
news articles from sections
you wouldn’t click on.
How can we find articles you might
like from sections that you wouldn’t
normally look at?
How can we find articles you might
like from sections that you wouldn’t
normally look at?
? ?
Place your screenshot here
InOtherNews.xyz
Data
Twitter API
47,000 network relations
Twitter and
Guardian APIs
1,300 Guardian
journalists
Guardian API
300 articles daily
(10,000+ total)
Transitivity
USER
JOURNALIST 1
JOURNALIST 2
likes to read
likes to read
might like to read?
How can transitivity help to solve our
problem?
USER
who likes
sport
SPORT
SPORT
SPORT
Narrow
horizons!
How can transitivity help to solve our
problem?
USER
who likes
sport
SPORT
World
SPORT
SPORT
Music
TV
Politics
Business
SportBroad
horizons!
User-based recommender algorithm
Initial recommendations
Weighted k-nearest
neighbors
User
input
User
feedback
Updated recommendations
Incorporating upvotes as
additional weighted
neighbors.
Journalist features
Neighbourhood component
analysis
User-based recommender algorithm
Leave-one-out cross validation
Initial recommendations
Weighted k-nearest
neighbors
User
input
Optimizing k gives 51%
more correct followees
than just recommending
most popular journalists.
Optimizing weights
gives up to 5% more
correct followees than
using method above
alone.
User
feedback
Updated recommendations
Incorporating upvotes as
additional weighted
neighbors.
Journalist features
Neighbourhood component
analysis
Charlotte Greenan
Homophily
Transitivity
12times as many
triangles as a random
graph
Data
◎ Articles from Guardian API.
◎ Social network from Twitter
API.
Leave-one-out cross validation
Improvement in correctly predicted ties by section
Neighbourhood component analysis
Transforming similarities between sections
Before After
Recommendation algorithm
◎ k most similar journalists, (cosine similarity);
◎ Journalists you like, (user feedback);
◎ Journalists you don’t like, (user feedback).
◎ Order journalists by their score:
◎ Recommend journalists using order until all
sections recommended (or score is zero).
Credits
Special thanks to all the people who made and released
these awesome resources for free:
◎ Simple line icons by Mirko Monti
◎ E-commerce icons by Virgil Pana
◎ Streamline iconset by Webalys
◎ Presentation template by SlidesCarnival
◎ Photographs by Unsplash & Death to the Stock Photo
(license)

More Related Content

Similar to Slides

Social Websites And Seo Social Dev Camp Chicago2008 By John Fairley
Social Websites And Seo Social Dev Camp Chicago2008 By John FairleySocial Websites And Seo Social Dev Camp Chicago2008 By John Fairley
Social Websites And Seo Social Dev Camp Chicago2008 By John FairleyJohn Fairley
 
ML&AI APPROACH TO USER UNDERSTANDING ECOSYSTEM AT VCCORP Applications to News...
ML&AI APPROACH TO USER UNDERSTANDING ECOSYSTEM AT VCCORP Applications to News...ML&AI APPROACH TO USER UNDERSTANDING ECOSYSTEM AT VCCORP Applications to News...
ML&AI APPROACH TO USER UNDERSTANDING ECOSYSTEM AT VCCORP Applications to News...Tuan Hoang
 
Advanced Keyword Research
Advanced Keyword ResearchAdvanced Keyword Research
Advanced Keyword ResearchDave Snyder
 
Intro to Web Optimization by Jennifer Lill
Intro to Web Optimization by Jennifer LillIntro to Web Optimization by Jennifer Lill
Intro to Web Optimization by Jennifer LillJennifer Lill
 
Best Free SEO Tools - digiayushman.com.pdf
Best Free SEO Tools - digiayushman.com.pdfBest Free SEO Tools - digiayushman.com.pdf
Best Free SEO Tools - digiayushman.com.pdfdigiayushman4
 
UserVoice Summit 2013 Fall Keynote
UserVoice Summit 2013 Fall KeynoteUserVoice Summit 2013 Fall Keynote
UserVoice Summit 2013 Fall KeynoteUserVoice
 
Final Year PPT on Twitter App
Final Year PPT on Twitter AppFinal Year PPT on Twitter App
Final Year PPT on Twitter Appscorpionking257
 
FSE 2016: Mobile analytics in context. Keynote Seattle Nov 14 2016
FSE 2016: Mobile analytics in context. Keynote Seattle Nov 14 2016FSE 2016: Mobile analytics in context. Keynote Seattle Nov 14 2016
FSE 2016: Mobile analytics in context. Keynote Seattle Nov 14 2016Ankit Jain
 
PeopleBrowsr Platform - Monitoring What's Hot
PeopleBrowsr Platform - Monitoring What's HotPeopleBrowsr Platform - Monitoring What's Hot
PeopleBrowsr Platform - Monitoring What's HotPeopleBrowsr
 
Twitch Insights: Player Research and Interaction
Twitch Insights: Player Research and InteractionTwitch Insights: Player Research and Interaction
Twitch Insights: Player Research and InteractionTwitch Developers
 
Definition of criteria_for_impact_analysis
Definition of criteria_for_impact_analysisDefinition of criteria_for_impact_analysis
Definition of criteria_for_impact_analysisSTIinnsbruck
 
Eyefortravel prague LikeCube
Eyefortravel prague LikeCubeEyefortravel prague LikeCube
Eyefortravel prague LikeCubeLikeCube
 
Eyefortravel prague like_cube
Eyefortravel prague like_cubeEyefortravel prague like_cube
Eyefortravel prague like_cubeEmmanuel Marchal
 
Analytics For Leaders
Analytics For LeadersAnalytics For Leaders
Analytics For LeadersTunheim
 
Improving Game Performance and Streaming with Twitch Insights
Improving Game Performance and Streaming with Twitch Insights Improving Game Performance and Streaming with Twitch Insights
Improving Game Performance and Streaming with Twitch Insights Twitch Developers
 
Optimize Your Bottom Line
Optimize Your Bottom LineOptimize Your Bottom Line
Optimize Your Bottom LineInflow
 

Similar to Slides (20)

Gio
GioGio
Gio
 
Social Websites And Seo Social Dev Camp Chicago2008 By John Fairley
Social Websites And Seo Social Dev Camp Chicago2008 By John FairleySocial Websites And Seo Social Dev Camp Chicago2008 By John Fairley
Social Websites And Seo Social Dev Camp Chicago2008 By John Fairley
 
ML&AI APPROACH TO USER UNDERSTANDING ECOSYSTEM AT VCCORP Applications to News...
ML&AI APPROACH TO USER UNDERSTANDING ECOSYSTEM AT VCCORP Applications to News...ML&AI APPROACH TO USER UNDERSTANDING ECOSYSTEM AT VCCORP Applications to News...
ML&AI APPROACH TO USER UNDERSTANDING ECOSYSTEM AT VCCORP Applications to News...
 
Advanced Keyword Research
Advanced Keyword ResearchAdvanced Keyword Research
Advanced Keyword Research
 
Intro to Web Optimization by Jennifer Lill
Intro to Web Optimization by Jennifer LillIntro to Web Optimization by Jennifer Lill
Intro to Web Optimization by Jennifer Lill
 
Best Free SEO Tools - digiayushman.com.pdf
Best Free SEO Tools - digiayushman.com.pdfBest Free SEO Tools - digiayushman.com.pdf
Best Free SEO Tools - digiayushman.com.pdf
 
UserVoice Summit 2013 Fall Keynote
UserVoice Summit 2013 Fall KeynoteUserVoice Summit 2013 Fall Keynote
UserVoice Summit 2013 Fall Keynote
 
Final Year PPT on Twitter App
Final Year PPT on Twitter AppFinal Year PPT on Twitter App
Final Year PPT on Twitter App
 
FSE 2016: Mobile analytics in context. Keynote Seattle Nov 14 2016
FSE 2016: Mobile analytics in context. Keynote Seattle Nov 14 2016FSE 2016: Mobile analytics in context. Keynote Seattle Nov 14 2016
FSE 2016: Mobile analytics in context. Keynote Seattle Nov 14 2016
 
PeopleBrowsr Platform - Monitoring What's Hot
PeopleBrowsr Platform - Monitoring What's HotPeopleBrowsr Platform - Monitoring What's Hot
PeopleBrowsr Platform - Monitoring What's Hot
 
Twitch Insights: Player Research and Interaction
Twitch Insights: Player Research and InteractionTwitch Insights: Player Research and Interaction
Twitch Insights: Player Research and Interaction
 
Definition of criteria_for_impact_analysis
Definition of criteria_for_impact_analysisDefinition of criteria_for_impact_analysis
Definition of criteria_for_impact_analysis
 
Edge rank
Edge rankEdge rank
Edge rank
 
Eyefortravel prague LikeCube
Eyefortravel prague LikeCubeEyefortravel prague LikeCube
Eyefortravel prague LikeCube
 
Eyefortravel prague like_cube
Eyefortravel prague like_cubeEyefortravel prague like_cube
Eyefortravel prague like_cube
 
SEO PPT
SEO PPTSEO PPT
SEO PPT
 
Analytics For Leaders
Analytics For LeadersAnalytics For Leaders
Analytics For Leaders
 
Improving Game Performance and Streaming with Twitch Insights
Improving Game Performance and Streaming with Twitch Insights Improving Game Performance and Streaming with Twitch Insights
Improving Game Performance and Streaming with Twitch Insights
 
Optimize Your Bottom Line
Optimize Your Bottom LineOptimize Your Bottom Line
Optimize Your Bottom Line
 
Seo
SeoSeo
Seo
 

Slides

  • 1. In Other News Broaden your horizons by reading like a writer Charlotte Greenan Using a social network of journalists to recommend news articles from sections you wouldn’t click on.
  • 2. How can we find articles you might like from sections that you wouldn’t normally look at?
  • 3. How can we find articles you might like from sections that you wouldn’t normally look at? ? ?
  • 4. Place your screenshot here InOtherNews.xyz
  • 5. Data Twitter API 47,000 network relations Twitter and Guardian APIs 1,300 Guardian journalists Guardian API 300 articles daily (10,000+ total)
  • 6. Transitivity USER JOURNALIST 1 JOURNALIST 2 likes to read likes to read might like to read?
  • 7. How can transitivity help to solve our problem? USER who likes sport SPORT SPORT SPORT Narrow horizons!
  • 8. How can transitivity help to solve our problem? USER who likes sport SPORT World SPORT SPORT Music TV Politics Business SportBroad horizons!
  • 9. User-based recommender algorithm Initial recommendations Weighted k-nearest neighbors User input User feedback Updated recommendations Incorporating upvotes as additional weighted neighbors. Journalist features Neighbourhood component analysis
  • 10. User-based recommender algorithm Leave-one-out cross validation Initial recommendations Weighted k-nearest neighbors User input Optimizing k gives 51% more correct followees than just recommending most popular journalists. Optimizing weights gives up to 5% more correct followees than using method above alone. User feedback Updated recommendations Incorporating upvotes as additional weighted neighbors. Journalist features Neighbourhood component analysis
  • 12.
  • 13.
  • 14.
  • 15.
  • 16.
  • 17. Homophily Transitivity 12times as many triangles as a random graph Data ◎ Articles from Guardian API. ◎ Social network from Twitter API.
  • 18. Leave-one-out cross validation Improvement in correctly predicted ties by section
  • 19. Neighbourhood component analysis Transforming similarities between sections Before After
  • 20. Recommendation algorithm ◎ k most similar journalists, (cosine similarity); ◎ Journalists you like, (user feedback); ◎ Journalists you don’t like, (user feedback). ◎ Order journalists by their score: ◎ Recommend journalists using order until all sections recommended (or score is zero).
  • 21. Credits Special thanks to all the people who made and released these awesome resources for free: ◎ Simple line icons by Mirko Monti ◎ E-commerce icons by Virgil Pana ◎ Streamline iconset by Webalys ◎ Presentation template by SlidesCarnival ◎ Photographs by Unsplash & Death to the Stock Photo (license)