Successfully reported this slideshow.
We use your LinkedIn profile and activity data to personalize ads and to show you more relevant ads. You can change your ad preferences anytime.
Gaza War

Britches
World War II

Berlin Wall

Woodstock
1950

1900

1910

1970

1920

9/11

Gulf War

1930

1980

1940

19...
Entity Linking for a personalized timeline of historic events

•

Motivation

•

Method
•
•

Part II: Generate User Profile...
•

[…] To design and build innovative and robust prototypes and
demos for tools that analyse and/or integrate open web dat...
History education
Personalized historic timeline

Gaza War

Britches
World War II

Berlin Wall

Woodstock
1950

1900

1910

1970

1920

9/11...
Part I: Candidate Historic Events
Part I: Candidate Historic Events

select	
  ?concept	
  	
  
where	
  {	
  	
  
	
   ?concept	
  rdf:type	
  dbpedia-­‐ow...
concept	
  	
  
	
  
ept	
  rdf:type	
  dbpedia-­‐owl:Event	
  	
  
concept	
  	
  
	
  
ept	
  rdf:type	
  dbpedia-­‐owl:Event	
  	
  
Part II: User Profile

MY FACEBOOK
PROFILE

BIO

POST
POST

LIKES

POST
Extract Information from Facebook profile

MY FACEBOOK
PROFILE

BIO

POST
POST

LIKES

POST
Access Facebook profile

MY FACEBOOK
PROFILE

BIO

POST
POST

LIKES

POST

{	
  
"id":	
  "1183880085",	
  
"likes":	
  {	
...
Extract text
attributes

•
•
•
•
•
•

{	
  
"id":	
  "1183880085",	
  
"likes":	
  {	
  
	
  	
  	
  	
  "data":	
  [	
  
...
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•

ASAP	
  Rocky	
  
Ab-­‐Soul	
  
Chance	
  The	
  Rapper	
  
Canni...
Entity Linking
•

Given a Knowledge Base

•

Link mentions of entities (or concepts) to their referent entities
Entity Linking
•

From Wikipedia:
•

Extract anchor texts (words used to link to Wikipedia pages)
!
!
!
!
!
!

•

For each...
Entity Linking Example
Link Probability
“Nas” occurs 2475x in Wikipedia

!

is anchor

1.723x

is no anchor

752x
Entity Linking Example
Link Probability
“Nas” occurs 2475x in Wikipedia

!

is anchor

1723/2475

=

69,6%

is no anchor

...
Entity Linking Example
Commonness
•

Nas is used to refer to:
•

http://en.wikipedia.org/wiki/Nas

•

http://en.wikipedia....
Entity Linking Example
Commonness
•

Nas is used to refer to:
•

http://en.wikipedia.org/wiki/Nas

14x

•

http://en.wikip...
Entity Linking Example
Commonness
•

Nas is used to refer to:
•

http://en.wikipedia.org/wiki/Nas

14/25 =

56%

•

http:/...
{	
  
	
  	
  	
  	
  "text":	
  "Nas",	
  
	
  	
  	
  	
  "links":	
  [	
  
	
  	
  	
  	
  	
  	
  	
  	
  {	
  
	
  	
...
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•

AT5	
  
Mad	
  Men	
  
The	
  Wire	
  
Monty	
  Python's	
  Flying	
  
Circus	
...
Match Events to Profile Entities
Match Events to Profile Entities
Map Events to Wikipedia Entities
Match Events to Profile Entities
Matching metric #1: link overlap
Matching metric #1: link overlap
U.S.

Hiphop

NAS

Kanye!
West

Jay-Z
Damian!
Marley
Global!
War

U.S.
U.S.
Allies
Hiphop

Axis

NAS

Kanye!
West

Jay-Z
Damian!
Marley

World!
War II
Global!
War

U.S.
U.S.
Allies
Hiphop

Axis

NAS

Kanye!
West

Jay-Z
Damian!
Marley

1
World!
War II
Global!
War

1

U.S.

World!
War II

U.S.
Allies
Hiphop

Axis

NAS

Kanye!
West

Jay-Z
Damian!
Marley

Jay-Z

Hiphop

Kany...
Global!
War

1

U.S.

World!
War II

U.S.
Allies
Hiphop

Axis

NAS

Kanye!
West

Jay-Z
Damian!
Marley

Jay-Z

Hiphop

Kany...
Matching metric #2: direct link

U.S.

Hiphop

NAS

Kanye!
West

Jay-Z
Damian!
Marley

Jay-Z

Hiphop

Kanye!
West

51st!
G...
Matching metric #3: textual similarity
NAS

51st!
Grammy!
Awards
Matching metric #3: textual similarity
NAS

51st!
Grammy!
Awards
Matching metric #3: textual similarity
NAS

51st!
Grammy!
Awards
51st!
Grammy!
Awards

World!
War II

Score: 0.74

Score: 0.35
Combine scores & rank events
	
  	
  	
  	
  "5043324":	
  {	
  
	
  	
  	
  	
  	
  	
  "event_title":	
  "Iraq	
  War",	...
Future Work
•

Log interactions

•

Interpret clicks as (implicit) feedback:
•

Click on Event: user is interested

•

No ...
Thank you! Questions?
Try yourHistory:
See our poster:

http://apps.facebook.com/yourHistory

#98

!
!
!
!






David Gra...
yourHistory - entity linking for a personalized timeline of historic events
yourHistory - entity linking for a personalized timeline of historic events
yourHistory - entity linking for a personalized timeline of historic events
yourHistory - entity linking for a personalized timeline of historic events
yourHistory - entity linking for a personalized timeline of historic events
yourHistory - entity linking for a personalized timeline of historic events
yourHistory - entity linking for a personalized timeline of historic events
yourHistory - entity linking for a personalized timeline of historic events
yourHistory - entity linking for a personalized timeline of historic events
yourHistory - entity linking for a personalized timeline of historic events
yourHistory - entity linking for a personalized timeline of historic events
yourHistory - entity linking for a personalized timeline of historic events
Upcoming SlideShare
Loading in …5
×

yourHistory - entity linking for a personalized timeline of historic events

3,432 views

Published on

slides for yH talk @ ICT.OPEN2013 (Intelligent Systems track)

  • Be the first to comment

  • Be the first to like this

yourHistory - entity linking for a personalized timeline of historic events

  1. 1. Gaza War Britches World War II Berlin Wall Woodstock 1950 1900 1910 1970 1920 9/11 Gulf War 1930 1980 1940 1950 1990 1960 BET Hiphop Awards 2000 1970 1980 2010 1990 2000 David Graus, Maria-Hendrike Peetz, Daan Odijk, Maarten de Rijke, Ork de Rooij 2010
  2. 2. Entity Linking for a personalized timeline of historic events • Motivation • Method • • Part II: Generate User Profile • Part III: Matching Events to User Profile • • Part I: Fetch Candidate Historic Events Part IV: Scoring & Ranking Events Future Work
  3. 3. • […] To design and build innovative and robust prototypes and demos for tools that analyse and/or integrate open web data for educational purposes.
  4. 4. History education
  5. 5. Personalized historic timeline Gaza War Britches World War II Berlin Wall Woodstock 1950 1900 1910 1970 1920 9/11 Gulf War 1930 1980 1940 1950 1990 1960 BET Hiphop Awards 2000 1970 1980 2010 1990 2000 2010
  6. 6. Part I: Candidate Historic Events
  7. 7. Part I: Candidate Historic Events select  ?concept     where  {       ?concept  rdf:type  dbpedia-­‐owl:Event       }
  8. 8. concept       ept  rdf:type  dbpedia-­‐owl:Event    
  9. 9. concept       ept  rdf:type  dbpedia-­‐owl:Event    
  10. 10. Part II: User Profile MY FACEBOOK PROFILE BIO POST POST LIKES POST
  11. 11. Extract Information from Facebook profile MY FACEBOOK PROFILE BIO POST POST LIKES POST
  12. 12. Access Facebook profile MY FACEBOOK PROFILE BIO POST POST LIKES POST {   "id":  "1183880085",   "likes":  {          "data":  [              {                  "category":  "Musician/band",                  "created_time":  "2013-­‐10-­‐27T11:37:51+0                "name":  "NAS",                  "id":  "113591595350795"              },              {                  "category":  "Company",                  "created_time":  "2013-­‐10-­‐17T07:45:36+0                "name":  "Infinibase",                  "id":  "573216229380347"              },              {                  "category":  "Magazine",                  "created_time":  "2013-­‐10-­‐04T13:55:10+0                "name":  "New  Scientist  NL",                  "id":  "369158433181445"              },  
  13. 13. Extract text attributes • • • • • • {   "id":  "1183880085",   "likes":  {          "data":  [              {                  "category":  "Musician/band",                  "created_time":  "2013-­‐10-­‐27T11:37:51+0000",                  "name":  "NAS",                  "id":  "113591595350795"              },              {                  "category":  "Company",                  "created_time":  "2013-­‐10-­‐17T07:45:36+0000",                  "name":  "Infinibase",                  "id":  "573216229380347"              },              {                  "category":  "Magazine",                  "created_time":  "2013-­‐10-­‐04T13:55:10+0000",                  "name":  "New  Scientist  NL",                  "id":  "369158433181445"              },              {                  "category":  "Tv  show",                  "created_time":  "2010-­‐05-­‐09T01:06:27+0000",                  "name":  "The  Wire",                  "id":  "5991693871"              }  ]   } • • • • • • • • • • • • • • Story   Omroep  Maxim   Gamer01   Breaking  Bad   AT5   Mad  Men   The  Wire   Monty  Python's   Flying  Circus   Flight  of  the   Conchords   Donnie  Darko   Flevopark  Film   Festival   Do  The  Right   Thing   A  Clockwork   Orange   Wild  Style   Princess   Mononoke   The  Fountain   Pi   Northfork   La  Haine   Zen  and  the  Art   of  Motorcycle   Maintenance   Moon  Palace   • • • • • • • • • • • • • • • • • • • • • • • • Fountainhead   The  Wind-­‐Up   Bird  Chronicle   Wu-­‐Tang   J.Cole   NAS   Pusha  T   ASAP  Rocky   Ab-­‐Soul   Chance  The   Rapper   Cannibal  Ox   Bonobo   Aesop  Rock   Boards  Of   Canada   Jurassic  5   GREMS   Quasimoto   Strange  Journey   Volume  Three   Drop  Velvet   MODESELEKTOR   IAM   Derek   The  Onion   Imgur   De  Speld   Wu-­‐Tang  
  14. 14. • • • • • • • • • • • • • • • • • • • • • • • • • • • • ASAP  Rocky   Ab-­‐Soul   Chance  The  Rapper   Cannibal  Ox   Bonobo   Aesop  Rock   Boards  Of  Canada   Jurassic  5   GREMS   Quasimoto   Strange  Journey  Volume  Three   Drop  Velvet   MODESELEKTOR   IAM   Derek   The  Onion   Imgur   De  Speld   Wu-­‐Tang   J.Cole   I  Am  Fucking  Ambivalent  About   Science   NAS   Pusha  T   ASAP  Rocky   Chrietitie   Infinibase   Marktplaatspoxc3xabzie   Jeannette  Span  :  Spelen  
  15. 15. Entity Linking • Given a Knowledge Base • Link mentions of entities (or concepts) to their referent entities
  16. 16. Entity Linking • From Wikipedia: • Extract anchor texts (words used to link to Wikipedia pages) ! ! ! ! ! ! • For each n-gram n ↔ Wikipedia page W estimate: • Probability of using n-gram n to refer to Wikipedia page W
  17. 17. Entity Linking Example Link Probability “Nas” occurs 2475x in Wikipedia ! is anchor 1.723x is no anchor 752x
  18. 18. Entity Linking Example Link Probability “Nas” occurs 2475x in Wikipedia ! is anchor 1723/2475 = 69,6% is no anchor 752/2475 = 30.4%
  19. 19. Entity Linking Example Commonness • Nas is used to refer to: • http://en.wikipedia.org/wiki/Nas • http://en.wikipedia.org/wiki/Naas • http://en.wikipedia.org/wiki/Nås • http://en.wikipedia.org/wiki/Nas (Ikaria) • http://en.wikipedia.org/wiki/Untitled Nas album
  20. 20. Entity Linking Example Commonness • Nas is used to refer to: • http://en.wikipedia.org/wiki/Nas 14x • http://en.wikipedia.org/wiki/Naas 4x • http://en.wikipedia.org/wiki/Nås 3x • http://en.wikipedia.org/wiki/Nas (Ikaria) 2x • http://en.wikipedia.org/wiki/Untitled Nas album 2x
  21. 21. Entity Linking Example Commonness • Nas is used to refer to: • http://en.wikipedia.org/wiki/Nas 14/25 = 56% • http://en.wikipedia.org/wiki/Naas 4/25 = 1.6% • http://en.wikipedia.org/wiki/Nås 3/25 = 1.2% • http://en.wikipedia.org/wiki/Nas (Ikaria) 2/25 = 0.8% • http://en.wikipedia.org/wiki/Untitled Nas album 2/25 = 0.8%
  22. 22. {          "text":  "Nas",          "links":  [                  {                          "senseProbability":  0.726027397260274,                          "title":  "Nas",                          "url":  "http://en.wikipedia.org/wiki/Nas"                  },                  {                          "senseProbability":  0.125,                          "title":  "Naas",                          "url":  "http://en.wikipedia.org/wiki/Naas"                  },                  {                          "senseProbability":  0.1111111111111111,                          "title":  "Nås",                          "url":  "http://en.wikipedia.org/wiki/N%C3%A5s"                  },                  {                          "senseProbability":  0.0006523157208088715,                          "title":  "Nas  (Ikaria)",                          "url":  "http://en.wikipedia.org/wiki/Nas%20%28Ikaria%29"                  },                  {                          "senseProbability":  0.0006523157208088715,                          "title":  "Untitled  Nas  album",                          "url":  "http://en.wikipedia.org/wiki/Untitled%20Nas%20album"                  }   }
  23. 23. • • • • • • • • • • • • • • • • • • • • • AT5   Mad  Men   The  Wire   Monty  Python's  Flying   Circus   Flight  of  the  Conchords   Donnie  Darko   Flevopark  Film  Festival   Do  The  Right  Thing   A  Clockwork  Orange   Wild  Style   Princess  Mononoke   The  Fountain   Pi   Northfork   La  Haine   Zen  and  the  Art  of   Motorcycle  Maintenance   Moon  Palace   The  Fountainhead   The  Wind-­‐Up  Bird   Chronicle   Wu-­‐Tang   J.Cole  
  24. 24. Match Events to Profile Entities
  25. 25. Match Events to Profile Entities
  26. 26. Map Events to Wikipedia Entities
  27. 27. Match Events to Profile Entities
  28. 28. Matching metric #1: link overlap
  29. 29. Matching metric #1: link overlap
  30. 30. U.S. Hiphop NAS Kanye! West Jay-Z Damian! Marley
  31. 31. Global! War U.S. U.S. Allies Hiphop Axis NAS Kanye! West Jay-Z Damian! Marley World! War II
  32. 32. Global! War U.S. U.S. Allies Hiphop Axis NAS Kanye! West Jay-Z Damian! Marley 1 World! War II
  33. 33. Global! War 1 U.S. World! War II U.S. Allies Hiphop Axis NAS Kanye! West Jay-Z Damian! Marley Jay-Z Hiphop Kanye! West Link #4 51st! Grammy! Awards
  34. 34. Global! War 1 U.S. World! War II U.S. Allies Hiphop Axis NAS Kanye! West Jay-Z Damian! Marley Jay-Z Hiphop Kanye! West Link #4 3 51st! Grammy! Awards
  35. 35. Matching metric #2: direct link U.S. Hiphop NAS Kanye! West Jay-Z Damian! Marley Jay-Z Hiphop Kanye! West 51st! Grammy! Awards
  36. 36. Matching metric #3: textual similarity NAS 51st! Grammy! Awards
  37. 37. Matching metric #3: textual similarity NAS 51st! Grammy! Awards
  38. 38. Matching metric #3: textual similarity NAS 51st! Grammy! Awards
  39. 39. 51st! Grammy! Awards World! War II Score: 0.74 Score: 0.35
  40. 40. Combine scores & rank events        "5043324":  {              "event_title":  "Iraq  War",              "related_entity_title":  "The  Wire",              "score":  1.0,              "event_date":  "2003-­‐03-­‐20"          },          "1376628":  {              "event_title":  "Blankets  (comics)",              "related_entity_title":  "Princess  Mononoke",              "score":  0.11465851113504691,              "event_date":  "2003-­‐07-­‐23"          },          "15694206":  {              "event_title":  "2006  LG  Hockey  Games",              "related_entity_title":  "Reimersholme",              "score":  0.3467068139664613,              "event_date":  "2006-­‐04-­‐29"          },          "4861876":  {              "event_title":  "2005  UEFA  Champions  League  Final",              "related_entity_title":  "Istanbul",              "score":  1.0,              "event_date":  "2005-­‐05-­‐25"          },          "31966809":  {              "event_title":  "63rd  Primetime  Emmy  Awards",              "related_entity_title":  "Mad  Men",              "score":  0.04039278737569369,              "event_date":  "2011-­‐09-­‐18"          },
  41. 41. Future Work • Log interactions • Interpret clicks as (implicit) feedback: • Click on Event: user is interested • No click on Event: user is not • Learn scoring & ranking functions
  42. 42. Thank you! Questions? Try yourHistory: See our poster: http://apps.facebook.com/yourHistory
 #98 ! ! ! ! 

 

 David Graus

 d.p.graus@uva.nl @dvdgrs

×