BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Pruning and grafting syntactic trees for cross-lingual transfer ta
 sks - Edoardo Ponti\, TAL\, University of Cambridge
DTSTART:20180209T120000Z
DTEND:20180209T130000Z
UID:TALK99520@talks.cam.ac.uk
CONTACT:Andrew Caines
DESCRIPTION:Universal Dependencies is a framework for annotating syntactic
  trees consistently across languages to facilitate multilingual NLP and cr
 oss-lingual transfer. However\, trees of equivalent sentences might assume
  non-overlapping shapes because of inherent typological variation. In part
 icular\, this anisomorphism is driven by the variation in 1) morphological
  assets and 2) in clause-level constructions (such as polar questions\, pr
 edicative possession\, relative clauses\, etc.). In this work\, we demonst
 rate that reducing the level of anisomorphism yields consistent gains for 
 cross-lingual transfer tasks. First\, we show how measuring anisomorphism 
 improves the selection of the source in Dependency Parsing transfer. Secon
 d\, we put forth a method to preprocess source trees matching their shapes
  with target trees inspired by typological documentation. This yields impr
 ovements in the BLEU scores of syntax-based Neural Machine Translation fro
 m Arabic to Dutch\, and from Indonesian to Portuguese: we release these ne
 w datasets with the code. Our results indicate that the compatibility of t
 he shapes of syntactic trees is crucial for source selection and for boost
 ing cross-lingual transfer.
LOCATION:FW26\, Computer Laboratory
END:VEVENT
END:VCALENDAR