BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Semi-supervised learning for automatic conceptual property extract
 ion - Colin Kelly\, University of Cambridge
DTSTART:20120601T113000Z
DTEND:20120601T120000Z
UID:TALK38392@talks.cam.ac.uk
CONTACT:Ekaterina Kochmar
DESCRIPTION:For a given concrete noun concept\, humans are usually able to
  cite\nproperties (e.g.\, elephant is animal\, car has wheels) of that con
 cept\;\ncognitive psychologists have theorised that such properties are\nf
 undamental to understanding the abstract mental representation of\nconcept
 s in the brain. Consequently\, the ability to automatically\nextract such 
 properties would be of enormous benefit to the field of\nexperimental psyc
 hology. This paper investigates the use of\nsemi-supervised learning and s
 upport vector machines to automatically\nextract concept-relation-feature 
 triples from two large corpora\n(Wikipedia and UKWAC) for concrete noun co
 ncepts. Previous approaches\nhave relied on manually-generated rules and h
 and-crafted resources such\nas WordNet\; our method requires neither yet a
 chieves better performance\nthan these prior approaches\, measured both by
  comparison with a property\nnorm-derived gold standard as well as direct 
 human evaluation. Our\ntechnique performs particularly well on extracting 
 features relevant to\na given concept\, and suggests a number of promising
  areas for future focus.
LOCATION:FW26\, Computer Laboratory
END:VEVENT
END:VCALENDAR