BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Two Approaches to Grammar Induction: From Plain Text to Semantic S
 upervision - Omri Abend\, The Hebrew University of Jerusalem 
DTSTART:20130306T120000Z
DTEND:20130306T130000Z
UID:TALK43327@talks.cam.ac.uk
CONTACT:Ekaterina Kochmar
DESCRIPTION:Grammatical representation is in wide use in a variety of NLP 
 tasks such as\nMachine Translation and Textual Entailment. Semi-supervised
  and\nunsupervised approaches to grammar induction are increasingly being 
 used\,\nand offer applicative and theoretical advantages over their superv
 ised\ncounterparts. However\, it is still an open question what type of in
 put is\nsufficient to learn grammar in all its complexity. This question i
 s\ndirectly related to one of the central questions in language science:\n
 whether and how language can be acquired from experience. I will discuss\n
 two approaches to this question that I have been pursuing in my research.\
 n\nThe fully unsupervised approach uses only plain text as input\, treatin
 g\nlanguage as an ordered collection of semantically-void symbols. I will\
 npresent a number of works that effectively apply this minimalist approach
 \nto tasks at the syntax-semantics interface\, such as argument identifica
 tion\nand the classification of verbal arguments to cores and adjuncts.\n\
 nHowever\, despite their appeal\, performance of unsupervised models still
 \nlags considerably behind the state of the art. In the second part of the
 \ntalk I will discuss a complementary approach to grammar induction\, whic
 h\nalso assumes semantic corpus annotation as input. Concretely\, I will\n
 present UCCA (Universal Conceptual Cognitive Analysis) -- a novel scheme\n
 that provides a formal framework for semantic representation in general an
 d\nfor grammar induction in particular. UCCA covers many of the most impor
 tant\nelements and relations present in linguistic utterances\, including 
 the\nargument structure of various types of predicates and the linkage bet
 ween\nthem\, but confines itself to semantic distinctions. For instance\, 
 UCCA\nrepresents the similarity between “John made an appearance” and 
 “John\nappeared”\, disregarding their syntactic differences. I will al
 so touch on\nour current efforts for constructing a UCCA parser and for ap
 plying UCCA to\nstatistical machine translation. A UCCA-annotated corpus w
 ill be released\nduring 2013.\n\nJoint work with Ari Rappoport.
LOCATION:FW11\, Computer Laboratory
END:VEVENT
END:VCALENDAR