BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Recommending relevant citations using CoreSC and Argumentative Zon
 ing - Daniel Duma\, University of Edinburgh
DTSTART:20161111T120000Z
DTEND:20161111T130000Z
UID:TALK67194@talks.cam.ac.uk
CONTACT:Kris Cao
DESCRIPTION:Wouldn't it be helpful if your text editor automatically sugge
 sted papers that are contextually relevant to your work? We concern oursel
 ves with this task: we desire to recommend contextually relevant citations
  to the author of a paper. A number of rhetorical annotation schemes for a
 cademic articles have been developed over the years\, and it has often bee
 n suggested that they could find application in Information Retrieval scen
 arios such as this one. We investigate the usefulness for this task of two
  sentence-based\, functional\, scientific discourse annotation schemes: Co
 reSC for biomedical science (e.g. Hypothesis\, Method\, Result\, etc.) and
  Argumentative Zoning (AZ) for computational linguistics. We apply this bo
 th to the contents of articles and to anchor text\, that is\, the text sur
 rounding a citation\, in citing articles. This is an important source of d
 ata for building document representations. By annotating each sentence in 
 every document with CoreSC and AZ and indexing them separately by sentence
  class\, we aim to build a more useful vector-space representation of docu
 ments in our collection. Our results show consistent links between types o
 f citing sentences and types of cited sentences\, both in cited articles a
 nd in anchor text\, which we argue can indeed be exploited to increase the
  relevance of recommendations.\n
LOCATION:FW26\, Computer Laboratory
END:VEVENT
END:VCALENDAR