BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//talks.cam.ac.uk//v3//EN
BEGIN:VTIMEZONE
TZID:Europe/London
BEGIN:DAYLIGHT
TZOFFSETFROM:+0000
TZOFFSETTO:+0100
TZNAME:BST
DTSTART:19700329T010000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=-1SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:+0100
TZOFFSETTO:+0000
TZNAME:GMT
DTSTART:19701025T020000
RRULE:FREQ=YEARLY;BYMONTH=10;BYDAY=-1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
CATEGORIES:Language Technology Lab Seminars
SUMMARY:Natural Language Understanding and Generation with
  Abstract Meaning Representation - Marco Damonte\,
  the University of Edinburgh
DTSTART;TZID=Europe/London:20191121T110000
DTEND;TZID=Europe/London:20191121T120000
UID:TALK135109AThttp://talks.cam.ac.uk
URL:http://talks.cam.ac.uk/talk/index/135109
DESCRIPTION:In this talk\, I will discuss my recent work on pa
 rsing and generation with Abstract Meaning Represe
 ntation (AMR). AMR is a semantic representation fo
 r natural language that represents sentences as gr
 aphs\, where nodes represent concepts and edges re
 present semantic relations between them. Sentences
  are represented as graphs and not trees because n
 odes can have multiple incoming edges\, called ree
 ntrancies. These are due to several linguistic phe
 nomena such as control\, coreference\, and coordin
 ation. \n\nI will present my work on AMR parsing (
 from text to AMR) and AMR-to-text generation (from
  AMR to text). For the parsing task\, we showed th
 at it is possible to use techniques from tree pars
 ing and adapt them to parse AMR graphs. To better 
 analyze the quality of AMR parsers\, we developed 
 a set of fine-grained metrics to better assess par
 sers\, including a metric for reentrancy predictio
 n. Hence\, we performed a study of the main causes
  of reentrancies in AMR and their impact on perfor
 mance. For the generation task\, we showed that ne
 ural encoders that have access to reentrancies out
 perform those who do not\, demonstrating the impor
 tance of reentrancies also for generation.\n\nI wi
 ll also discuss the problem of using AMR for langu
 ages other than English. Annotating new AMR datase
 ts for other languages is an expensive process and
  requires defining ad-hoc annotation guidelines fo
 r each new language. It is therefore reasonable to
  ask whether we can share AMR annotations across l
 anguages.
LOCATION:Board room\, Faculty of English\, 9 West Rd (Sidgw
 ick Site)
CONTACT:Qianchu Liu
END:VEVENT
END:VCALENDAR