introduction to polimedia · henri beunders laura hollink damir juric geert jan houben jaap blom...
TRANSCRIPT
PoliMedia approach
PoliMedia Portal
Search debate and person
NewspapersKB
TelevisionSound and Vision
RadioKB
Dutch Digital
ParliamentKB
PoliMedia Link creation
Intuition 1: The news item contains a topic a/o name of a politician and is published within a week after a debate
Intuition 2: The more overlap in topics and named entities, the more probably there is a link.
“Give me all fragments of debates with over 60 related news items”
SELECT ?speech ?no_newsitems {{ SELECT ?speech (COUNT(?news) AS ?no_news_items)WHERE{
?speech <http://purl.org/linkedpolitics/nl/polivoc#coveredAt> ?news .
}GROUP BY ?speech }FILTER (?no_news_items > 60) }
SPARQL Endpoint
“Explore debates & media coverage”
PoliMedia.nl
• Yeah! It works (but no television)
• Not perfect, but still ok (recall: 62%; precision: 80%)
• It is open for everyone: www.polimedia.nl
• + via a Sparql Endpoint
• We won two prizes ☺
Results
• People actually use it ☺
• Demonstrates potential of linking data by using Natural Language Processing techniques
• National Library of the Netherlands explores possibilities to incorporate PoliMedia techniques in existing services
Results
• Can we incorporate more media: television, social media, websites?
• Is recall of 62% and precision of 80% enough for researchers?
• Sparql is horror for (most) Humanities researchers
• Imperfect data (OCR errors, lags in collection, …)
Open Issues
Credits
Martijn Kleppe
Max Kemman
Henri Beunders
Laura Hollink
Damir Juric
Geert Jan Houben
Jaap Blom
Johan Oomen
Financed by Data files