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.
Natural Language Processing
& Named Entity Recognition
marieke.van.erp@huygens.knaw.nl

merpeltje
Natural Language Processing
• Het verwerken van natuurlijke
(menselijke) taal door computers
• Doel: ‘begrijpen’ van tekst...
Natural Language Processing Stappen
Tokenisation
Lexical Analysis
Syntactic Analysis
Semantic Analysis
Pragmatic Analysis
...
Wat is Named Entity Recognition? 	
• Named Entity Recognition is het vinden en classificeren van namen in tekst
Abraham Lin...
Wat is Named Entity Recognition? 	
• Named Entity Recognition is het vinden en classificeren van namen in tekst
Abraham Lin...
Wat is Named Entity Recognition? 	
• Named Entity Recognition is het vinden en classificeren van namen in tekst
Abraham Lin...
Wat is Named Entity Recognition? 	
• Named Entity Recognition is het vinden en classificeren van namen in tekst
Abraham Lin...
Wat is Named Entity Recognition? 	
• Named Entity Recognition is het vinden en classificeren van namen in tekst
Abraham Lin...
Wat is Named Entity Recognition? 	
• Named Entity Recognition is het vinden en classificeren van namen in tekst
Abraham Lin...
Waarom zijn entiteiten zo belangrijk?
• Entiteiten kunnen worden geïndexeerd 

• Sentiment kan aan entiteiten gekoppeld
wo...
Entity recognition voor (literatuur)collecties
Entity recognition voor (literatuur)collecties
Hoe gaat Named Entity Recognition in zijn werk?
• Handgeschreven regels (bv. Persoonsnaam: “Meneer” gevolgd door
woord dat...
State-of-the-Art
*ACL: Association for Computational Linguistics
Valkuilen
• Hoge resultaten zijn uitsluitend op teksten ‘in domain’
teksten
• Meeste systemen zijn geoptimaliseerd voor he...
Als je er toch mee aan de slag wilt
• https://gate.ac.uk/
• https://nlp.stanford.edu/software/CRF-NER.shtml
• https://open...
Evalueer & vraag hulp
• Controleer of het NER systeem de gewenste output geeft
• Bijvoorbeeld door de output met de hand n...
Wat is Coreference Resolution
source: http://en.wikipedia.org/wiki/Abraham_Lincoln
• Coreference resolution is het vinden ...
Maar het is eigenlijk nog ingewikkelder
source: http://en.wikipedia.org/wiki/Abraham_Lincoln
• Coreference resolution is h...
Maar het is eigenlijk nog ingewikkelder
source: http://en.wikipedia.org/wiki/Abraham_Lincoln
• Coreference resolution is h...
Waarom is coreference resolutie belangrijk?
• Coreference is een frequent taalfenomeen 

• Coreference resolution is essen...
Intussen in Semantic Web-land
Grounding Entities in the LOD cloud
“Abraham Lincoln”
http://dbpedia.org/resource/Abraham_Lincoln
Waar het mis kan gaan
Paris
Waar het mis kan gaan
ParisImage source: https://static.pexels.com/photos/161853/eiffel-tower-paris-
france-tower-161853.j...
Waar het mis kan gaan
ParisImage source: https://static.pexels.com/photos/161853/eiffel-tower-paris-
france-tower-161853.j...
Waar het mis kan gaan
ParisImage source: https://static.pexels.com/photos/161853/eiffel-tower-paris-
france-tower-161853.j...
Waar het mis kan gaan
ParisImage source: https://static.pexels.com/photos/161853/eiffel-tower-paris-
france-tower-161853.j...
Waar het mis kan gaan
ParisImage source: https://static.pexels.com/photos/161853/eiffel-tower-paris-
france-tower-161853.j...
Waar het mis kan gaan
ParisImage source: https://static.pexels.com/photos/161853/eiffel-tower-paris-
france-tower-161853.j...
Maar je moet wel weten welke ‘Lincoln’ je moet hebben
Hoe gaat Named Entity Disambiguation in zijn werk?
• Er is een database van entiteiten met context
• Entiteiten worden geï...
State-of-the-Art
• Lehman et al. (2010): 80,6
• TAC KBP Entity Linking*: persoon, organisatie, locatie
• Nieuwsartikelen, ...
Valkuilen
• Vooral op ‘populaire’ entiteiten (beroemd genoeg voor
Wikipedia)
• Evaluatiedatasets vaak niet gebalanceerd (V...
Als je er zelf mee aan de slag wilt
• https://github.com/dbpedia-spotlight/dbpedia-spotlight/wiki
• http://nerd.eurecom.fr...
Evalueer & vraag hulp
• Controleer of het NEL systeem de gewenste output geeft
• Staan de entiteiten die voor mij relevant...
Recap
• Named entity recognition en disambiguatie zijn potentiële
verrijkingen van databronnen
• Vaak nog onderzoeksstadiu...
Verder lezen
• Lehmann, J., Monahan, S., Nezda, L., Jung, A., & Shi, Y. (2010). LCC
Approaches to Knowledge Base Populatio...
Img source: http://mwallacepro.com/content/uploads/2016/01/question-1500082_1920.jpg
Natural Language Processing en Named Entity Recognition
Natural Language Processing en Named Entity Recognition
Natural Language Processing en Named Entity Recognition
Natural Language Processing en Named Entity Recognition
Natural Language Processing en Named Entity Recognition
Natural Language Processing en Named Entity Recognition
Upcoming SlideShare
Loading in …5
×

Natural Language Processing en Named Entity Recognition

297 views

Published on

Slides van de presentatie gegeven bij de DH Clinics Universiteitsbibliotheken op 17 oktober 2017.
https://dhclinics.github.io/dag/2017-dag3

Published in: Technology
  • Be the first to comment

  • Be the first to like this

Natural Language Processing en Named Entity Recognition

  1. 1. Natural Language Processing & Named Entity Recognition marieke.van.erp@huygens.knaw.nl merpeltje
  2. 2. Natural Language Processing • Het verwerken van natuurlijke (menselijke) taal door computers • Doel: ‘begrijpen’ van tekst • In de praktijk: verrijken van collecties, verbeteren zoekmachines, automatisch vertalen • Begrijpen van tekst wordt voor computer opgedeeld in verschillende lagen/stapjes: Tokenisation Lexical Analysis Syntactic Analysis Semantic Analysis Pragmatic Analysis Input text Speaker's intended meaning
  3. 3. Natural Language Processing Stappen Tokenisation Lexical Analysis Syntactic Analysis Semantic Analysis Pragmatic Analysis Input text Speaker's intended meaning • Tokenisation: detectie van zinnen en woorden • Lexicale analyse: woordsoorten (voorzetsel, adjectief) en vervoegingen • Syntactische analyse: zinsstructuur (onderwerp, persoonsvorm, lijdend voorwerp) • Semantische analyse: naar wie wordt verwezen in de tekst, waar gaat de tekst eigenlijk over? • Pragmatische analyse: wat bedoelt de spreker
  4. 4. Wat is Named Entity Recognition? • Named Entity Recognition is het vinden en classificeren van namen in tekst Abraham Lincoln Listeni/ˈeɪbrəhæm ˈlɪŋkən/ (February 12, 1809 – April 15, 1865) was the 16th President of the United States, serving from March 1861 until his assassination in April 1865. Lincoln led the United States through its Civil War—its bloodiest war and its greatest moral, constitutional and political crisis.[1][2] In doing so, he preserved the Union, abolished slavery, strengthened the federal government, and modernized the economy. Lincoln grew up on the western frontier in Kentucky and Indiana. Largely self-educated, he became a lawyer in Illinois, a Whig Party leader, and a member of the Illinois House of Representatives, where he served from 1834 to 1846. Elected to the United States House of Representatives in 1846, Lincoln promoted rapid modernization of the economy through banks, tariffs, and railroads. Because he had originally agreed not to run for a second term in Congress, and his opposition to the Mexican–American War was unpopular among Illinois voters, Lincoln returned to Springfield and resumed his successful law practice. Reentering politics in 1854, he became a leader in building the new Republican Party, which had a statewide majority in Illinois. In 1858, while taking part in a series of highly publicized debates with his opponent and rival, Democrat Stephen A. Douglas, Lincoln spoke out against the expansion of slavery, but lost the U.S. Senate race to Douglas. source: http://en.wikipedia.org/wiki/Abraham_Lincoln
  5. 5. Wat is Named Entity Recognition? • Named Entity Recognition is het vinden en classificeren van namen in tekst Abraham Lincoln Listeni/ˈeɪbrəhæm ˈlɪŋkən/ (February 12, 1809 – April 15, 1865) was the 16th President of the United States, serving from March 1861 until his assassination in April 1865. Lincoln led the United States through its Civil War—its bloodiest war and its greatest moral, constitutional and political crisis.[1][2] In doing so, he preserved the Union, abolished slavery, strengthened the federal government, and modernized the economy. Lincoln grew up on the western frontier in Kentucky and Indiana. Largely self-educated, he became a lawyer in Illinois, a Whig Party leader, and a member of the Illinois House of Representatives, where he served from 1834 to 1846. Elected to the United States House of Representatives in 1846, Lincoln promoted rapid modernization of the economy through banks, tariffs, and railroads. Because he had originally agreed not to run for a second term in Congress, and his opposition to the Mexican–American War was unpopular among Illinois voters, Lincoln returned to Springfield and resumed his successful law practice. Reentering politics in 1854, he became a leader in building the new Republican Party, which had a statewide majority in Illinois. In 1858, while taking part in a series of highly publicized debates with his opponent and rival, Democrat Stephen A. Douglas, Lincoln spoke out against the expansion of slavery, but lost the U.S. Senate race to Douglas. source: http://en.wikipedia.org/wiki/Abraham_Lincoln
  6. 6. Wat is Named Entity Recognition? • Named Entity Recognition is het vinden en classificeren van namen in tekst Abraham Lincoln Listeni/ˈeɪbrəhæm ˈlɪŋkən/ (February 12, 1809 – April 15, 1865) was the 16th President of the United States, serving from March 1861 until his assassination in April 1865. Lincoln led the United States through its Civil War—its bloodiest war and its greatest moral, constitutional and political crisis.[1][2] In doing so, he preserved the Union, abolished slavery, strengthened the federal government, and modernized the economy. Lincoln grew up on the western frontier in Kentucky and Indiana. Largely self-educated, he became a lawyer in Illinois, a Whig Party leader, and a member of the Illinois House of Representatives, where he served from 1834 to 1846. Elected to the United States House of Representatives in 1846, Lincoln promoted rapid modernization of the economy through banks, tariffs, and railroads. Because he had originally agreed not to run for a second term in Congress, and his opposition to the Mexican–American War was unpopular among Illinois voters, Lincoln returned to Springfield and resumed his successful law practice. Reentering politics in 1854, he became a leader in building the new Republican Party, which had a statewide majority in Illinois. In 1858, while taking part in a series of highly publicized debates with his opponent and rival, Democrat Stephen A. Douglas, Lincoln spoke out against the expansion of slavery, but lost the U.S. Senate race to Douglas. source: http://en.wikipedia.org/wiki/Abraham_Lincoln
  7. 7. Wat is Named Entity Recognition? • Named Entity Recognition is het vinden en classificeren van namen in tekst Abraham Lincoln Listeni/ˈeɪbrəhæm ˈlɪŋkən/ (February 12, 1809 – April 15, 1865) was the 16th President of the United States, serving from March 1861 until his assassination in April 1865. Lincoln led the United States through its Civil War—its bloodiest war and its greatest moral, constitutional and political crisis.[1][2] In doing so, he preserved the Union, abolished slavery, strengthened the federal government, and modernized the economy. Lincoln grew up on the western frontier in Kentucky and Indiana. Largely self-educated, he became a lawyer in Illinois, a Whig Party leader, and a member of the Illinois House of Representatives, where he served from 1834 to 1846. Elected to the United States House of Representatives in 1846, Lincoln promoted rapid modernization of the economy through banks, tariffs, and railroads. Because he had originally agreed not to run for a second term in Congress, and his opposition to the Mexican–American War was unpopular among Illinois voters, Lincoln returned to Springfield and resumed his successful law practice. Reentering politics in 1854, he became a leader in building the new Republican Party, which had a statewide majority in Illinois. In 1858, while taking part in a series of highly publicized debates with his opponent and rival, Democrat Stephen A. Douglas, Lincoln spoke out against the expansion of slavery, but lost the U.S. Senate race to Douglas. source: http://en.wikipedia.org/wiki/Abraham_Lincoln
  8. 8. Wat is Named Entity Recognition? • Named Entity Recognition is het vinden en classificeren van namen in tekst Abraham Lincoln Listeni/ˈeɪbrəhæm ˈlɪŋkən/ (February 12, 1809 – April 15, 1865) was the 16th President of the United States, serving from March 1861 until his assassination in April 1865. Lincoln led the United States through its Civil War—its bloodiest war and its greatest moral, constitutional and political crisis.[1][2] In doing so, he preserved the Union, abolished slavery, strengthened the federal government, and modernized the economy. Lincoln grew up on the western frontier in Kentucky and Indiana. Largely self-educated, he became a lawyer in Illinois, a Whig Party leader, and a member of the Illinois House of Representatives, where he served from 1834 to 1846. Elected to the United States House of Representatives in 1846, Lincoln promoted rapid modernization of the economy through banks, tariffs, and railroads. Because he had originally agreed not to run for a second term in Congress, and his opposition to the Mexican–American War was unpopular among Illinois voters, Lincoln returned to Springfield and resumed his successful law practice. Reentering politics in 1854, he became a leader in building the new Republican Party, which had a statewide majority in Illinois. In 1858, while taking part in a series of highly publicized debates with his opponent and rival, Democrat Stephen A. Douglas, Lincoln spoke out against the expansion of slavery, but lost the U.S. Senate race to Douglas. source: http://en.wikipedia.org/wiki/Abraham_Lincoln
  9. 9. Wat is Named Entity Recognition? • Named Entity Recognition is het vinden en classificeren van namen in tekst Abraham Lincoln Listeni/ˈeɪbrəhæm ˈlɪŋkən/ (February 12, 1809 – April 15, 1865) was the 16th President of the United States, serving from March 1861 until his assassination in April 1865. Lincoln led the United States through its Civil War—its bloodiest war and its greatest moral, constitutional and political crisis.[1][2] In doing so, he preserved the Union, abolished slavery, strengthened the federal government, and modernized the economy. Lincoln grew up on the western frontier in Kentucky and Indiana. Largely self-educated, he became a lawyer in Illinois, a Whig Party leader, and a member of the Illinois House of Representatives, where he served from 1834 to 1846. Elected to the United States House of Representatives in 1846, Lincoln promoted rapid modernization of the economy through banks, tariffs, and railroads. Because he had originally agreed not to run for a second term in Congress, and his opposition to the Mexican–American War was unpopular among Illinois voters, Lincoln returned to Springfield and resumed his successful law practice. Reentering politics in 1854, he became a leader in building the new Republican Party, which had a statewide majority in Illinois. In 1858, while taking part in a series of highly publicized debates with his opponent and rival, Democrat Stephen A. Douglas, Lincoln spoke out against the expansion of slavery, but lost the U.S. Senate race to Douglas. source: http://en.wikipedia.org/wiki/Abraham_Lincoln
  10. 10. Waarom zijn entiteiten zo belangrijk? • Entiteiten kunnen worden geïndexeerd • Sentiment kan aan entiteiten gekoppeld worden • Entiteiten vormen het startpunt voor veel andere informatie-extractietaken bron: https://trends.google.nl/trends/topcharts
  11. 11. Entity recognition voor (literatuur)collecties
  12. 12. Entity recognition voor (literatuur)collecties
  13. 13. Hoe gaat Named Entity Recognition in zijn werk? • Handgeschreven regels (bv. Persoonsnaam: “Meneer” gevolgd door woord dat begint met hoofdletter) • Hoge precisie, maar veel werk • Wat te doen met woorden die ambigu zijn? Apple, Handel, De Leeuw • Statistische (machine learning) methodes • Veel geannoteerde training data nodig • Semi-supervised machine learning wordt onderzocht om dit op te lossen
  14. 14. State-of-the-Art *ACL: Association for Computational Linguistics
  15. 15. Valkuilen • Hoge resultaten zijn uitsluitend op teksten ‘in domain’ teksten • Meeste systemen zijn geoptimaliseerd voor hedendaagse kranten • Taalvariatie (tijd, dialect, status, type publicatie) beïnvloeden resultaten • Advies: do not try this at home
  16. 16. Als je er toch mee aan de slag wilt • https://gate.ac.uk/ • https://nlp.stanford.edu/software/CRF-NER.shtml • https://opennlp.apache.org/ • http://ixa2.si.ehu.es/ ixa-pipes/ • https://languagemachines.github.io/frog/
  17. 17. Evalueer & vraag hulp • Controleer of het NER systeem de gewenste output geeft • Bijvoorbeeld door de output met de hand na te lopen • Schakel een taaltechnoloog in! • Computational Linguistics in the Netherlands • http://www.let.rug.nl/vannoord/Clin/
  18. 18. Wat is Coreference Resolution source: http://en.wikipedia.org/wiki/Abraham_Lincoln • Coreference resolution is het vinden van welke woorden/frasen naar hetzelfde object verwijzen Abraham Lincoln Listeni/ˈeɪbrəhæm ˈlɪŋkən/ (February 12, 1809 – April 15, 1865) was the 16th President of the United States, serving from March 1861 until his assassination in April 1865. Lincoln led the United States through its Civil War—its bloodiest war and its greatest moral, constitutional and political crisis.[1][2] In doing so, he preserved the Union, abolished slavery, strengthened the federal government, and modernized the economy. Lincoln grew up on the western frontier in Kentucky and Indiana. Largely self-educated, he became a lawyer in Illinois, a Whig Party leader, and a member of the Illinois House of Representatives, where he served from 1834 to 1846. Elected to the United States House of Representatives in 1846, Lincoln promoted rapid modernization of the economy through banks, tariffs, and railroads. Because he had originally agreed not to run for a second term in Congress, and his opposition to the Mexican–American War was unpopular among Illinois voters, Lincoln returned to Springfield and resumed his successful law practice. Reentering politics in 1854, he became a leader in building the new Republican Party, which had a statewide majority in Illinois. In 1858, while taking part in a series of highly publicized debates with his opponent and rival, Democrat Stephen A. Douglas, Lincoln spoke out against the expansion of slavery, but lost the U.S. Senate race to Douglas.
  19. 19. Maar het is eigenlijk nog ingewikkelder source: http://en.wikipedia.org/wiki/Abraham_Lincoln • Coreference resolution is het vinden van welke woorden/frasen naar hetzelfde object verwijzen Abraham Lincoln Listeni/ˈeɪbrəhæm ˈlɪŋkən/ (February 12, 1809 – April 15, 1865) was the 16th President of the United States, serving from March 1861 until his assassination in April 1865. Lincoln led the United States through its Civil War—its bloodiest war and its greatest moral, constitutional and political crisis.[1][2] In doing so, he preserved the Union, abolished slavery, strengthened the federal government, and modernized the economy. Lincoln grew up on the western frontier in Kentucky and Indiana. Largely self-educated, he became a lawyer in Illinois, a Whig Party leader, and a member of the Illinois House of Representatives, where he served from 1834 to 1846. Elected to the United States House of Representatives in 1846, Lincoln promoted rapid modernization of the economy through banks, tariffs, and railroads. Because he had originally agreed not to run for a second term in Congress, and his opposition to the Mexican–American War was unpopular among Illinois voters, Lincoln returned to Springfield and resumed his successful law practice. Reentering politics in 1854, he became a leader in building the new Republican Party, which had a statewide majority in Illinois. In 1858, while taking part in a series of highly publicized debates with his opponent and rival, Democrat Stephen A. Douglas, Lincoln spoke out against the expansion of slavery, but lost the U.S. Senate race to Douglas.
  20. 20. Maar het is eigenlijk nog ingewikkelder source: http://en.wikipedia.org/wiki/Abraham_Lincoln • Coreference resolution is het vinden van welke woorden/frasen naar hetzelfde object verwijzen Abraham Lincoln Listeni/ˈeɪbrəhæm ˈlɪŋkən/ (February 12, 1809 – April 15, 1865) was the 16th President of the United States, serving from March 1861 until his assassination in April 1865. Lincoln led the United States through its Civil War—its bloodiest war and its greatest moral, constitutional and political crisis.[1][2] In doing so, he preserved the Union, abolished slavery, strengthened the federal government, and modernized the economy. Lincoln grew up on the western frontier in Kentucky and Indiana. Largely self-educated, he became a lawyer in Illinois, a Whig Party leader, and a member of the Illinois House of Representatives, where he served from 1834 to 1846. Elected to the United States House of Representatives in 1846, Lincoln promoted rapid modernization of the economy through banks, tariffs, and railroads. Because he had originally agreed not to run for a second term in Congress, and his opposition to the Mexican–American War was unpopular among Illinois voters, Lincoln returned to Springfield and resumed his successful law practice. Reentering politics in 1854, he became a leader in building the new Republican Party, which had a statewide majority in Illinois. In 1858, while taking part in a series of highly publicized debates with his opponent and rival, Democrat Stephen A. Douglas, Lincoln spoke out against the expansion of slavery, but lost the U.S. Senate race to Douglas.
  21. 21. Waarom is coreference resolutie belangrijk? • Coreference is een frequent taalfenomeen • Coreference resolution is essentieel om een tekst goed te begrijpen • Coreference resolution is moeilijk omdat het over de grenzen van syntax, semantiek en discours gaat • Huidige state-of-the-art: 60~70% accuraat
  22. 22. Intussen in Semantic Web-land
  23. 23. Grounding Entities in the LOD cloud “Abraham Lincoln” http://dbpedia.org/resource/Abraham_Lincoln
  24. 24. Waar het mis kan gaan Paris
  25. 25. Waar het mis kan gaan ParisImage source: https://static.pexels.com/photos/161853/eiffel-tower-paris- france-tower-161853.jpeg
  26. 26. Waar het mis kan gaan ParisImage source: https://static.pexels.com/photos/161853/eiffel-tower-paris- france-tower-161853.jpeg https://upload.wikimedia.org/ wikipedia/commons/5/58/ Paris_Hilton_2%2C_2012.jpg
  27. 27. Waar het mis kan gaan ParisImage source: https://static.pexels.com/photos/161853/eiffel-tower-paris- france-tower-161853.jpeg https://upload.wikimedia.org/ wikipedia/commons/5/58/ Paris_Hilton_2%2C_2012.jpg https://commons.wikimedia.org/ wiki/File:Helene_Paris_David.jpg
  28. 28. Waar het mis kan gaan ParisImage source: https://static.pexels.com/photos/161853/eiffel-tower-paris- france-tower-161853.jpeg https://upload.wikimedia.org/ wikipedia/commons/5/58/ Paris_Hilton_2%2C_2012.jpg https://commons.wikimedia.org/ wiki/File:Helene_Paris_David.jpg https://upload.wikimedia.org/wikipedia/commons/4/49/ Eiffel_Tower_Replica_Paris_Texas_DSC_0602_ad.JPG
  29. 29. Waar het mis kan gaan ParisImage source: https://static.pexels.com/photos/161853/eiffel-tower-paris- france-tower-161853.jpeg https://upload.wikimedia.org/ wikipedia/commons/5/58/ Paris_Hilton_2%2C_2012.jpg https://commons.wikimedia.org/ wiki/File:Helene_Paris_David.jpg https://upload.wikimedia.org/wikipedia/commons/4/49/ Eiffel_Tower_Replica_Paris_Texas_DSC_0602_ad.JPG
  30. 30. Waar het mis kan gaan ParisImage source: https://static.pexels.com/photos/161853/eiffel-tower-paris- france-tower-161853.jpeg https://upload.wikimedia.org/ wikipedia/commons/5/58/ Paris_Hilton_2%2C_2012.jpg https://commons.wikimedia.org/ wiki/File:Helene_Paris_David.jpg https://upload.wikimedia.org/wikipedia/commons/4/49/ Eiffel_Tower_Replica_Paris_Texas_DSC_0602_ad.JPG
  31. 31. Maar je moet wel weten welke ‘Lincoln’ je moet hebben
  32. 32. Hoe gaat Named Entity Disambiguation in zijn werk? • Er is een database van entiteiten met context • Entiteiten worden geïdentificeerd in tekst • Het systeem zoekt per gevonden entiteit in de tekst een passende entiteit in de database • Matching op woorden • Matching op context • Ranking (vaak naar populariteit van entiteit in de database)
  33. 33. State-of-the-Art • Lehman et al. (2010): 80,6 • TAC KBP Entity Linking*: persoon, organisatie, locatie • Nieuwsartikelen, blogs en fora (Usenet) * entity linking en entity disambiguation worden vaak door elkaar gebruikt
  34. 34. Valkuilen • Vooral op ‘populaire’ entiteiten (beroemd genoeg voor Wikipedia) • Evaluatiedatasets vaak niet gebalanceerd (Van Erp et al. 2016) • Moeilijk in te richten voor eigen thesauri
  35. 35. Als je er zelf mee aan de slag wilt • https://github.com/dbpedia-spotlight/dbpedia-spotlight/wiki • http://nerd.eurecom.fr • FOX http://aksw.org/Projects/FOX.html • https://www.ibm.com/watson/services/natural- language-understanding/ • http://babelfy.org/
  36. 36. Evalueer & vraag hulp • Controleer of het NEL systeem de gewenste output geeft • Staan de entiteiten die voor mij relevant zijn wel in de knowledge base? • Schakel een Semantic Web expert in! • http://nl.dbpedia.org/web/ (Nederlandstalige DBpedia ) • https://iswc2017.semanticweb.org/ • Platform Linked Data Nederland: http://www.pilod.nl/ • https://2017.semantics.cc/
  37. 37. Recap • Named entity recognition en disambiguatie zijn potentiële verrijkingen van databronnen • Vaak nog onderzoeksstadium • Domeinadaptatie blijft een punt van aandacht • Maar er is wel veel interesse in taaltechnologie en semantic web gemeenschap om dit te verbeteren!
  38. 38. Verder lezen • Lehmann, J., Monahan, S., Nezda, L., Jung, A., & Shi, Y. (2010). LCC Approaches to Knowledge Base Population at TAC 2010. In TAC. • van Erp, M., Mendes, P. N., Paulheim, H., Ilievski, F., Plu, J., Rizzo, G., & Waitelonis, J. (2016). Evaluating Entity Linking: An Analysis of Current Benchmark Datasets and a Roadmap for Doing a Better Job. In LREC (Vol. 5, p. 2016). • Maynard, D., Bontcheva, K., & Augenstein, I. (2016) Natural Language Processing for the Semantic Web. Morgan Claypool. • Introductiefilmpje Semantic Web: https://www.youtube.com/watch? v=TJfrNo3Z-DU (Metaweb is opgekocht door Google, wat misschien iets te maken heeft met de gestructureerde zoekresultaten die je nu terugkrijgt die Seth noemde)
  39. 39. Img source: http://mwallacepro.com/content/uploads/2016/01/question-1500082_1920.jpg

×