BEGIN:VCALENDAR
VERSION:2.0
PRODID:icalendar-ruby
CALSCALE:GREGORIAN
X-WR-CALNAME:PhD Preliminary Oral Exam – Rasha Obeidat
X-WR-TIMEZONE:Pacific Time (US & Canada)
BEGIN:VEVENT
DTSTAMP:20260616T090633Z
UID:tag:localist.com\,2008:EventInstance_4374802
DTSTART:20190222T210000Z
DTEND:20190222T230000Z
DESCRIPTION:Learning with Limited Labeled Data in Natural Language Processi
 ng\n\nThe advent of deep learning models leads to a substantial improvemen
 t in a wide range of NLP tasks\, achieving state-of-art performances witho
 ut any hand-crafted features. However\, training deep models requires a ma
 ssive amount of labeled data. Labeling new data as a new task or domain em
 erges consumes time and efforts and needs domain expertise. As a result\, 
 the approaches that address the data scarcity are getting increasing atten
 tion in recent years\, including\, but not limited to\, transfer learning\
 , zero-shot learning\, and weak supervision. In this report\, we summarize
  our two prior works on learning from limited data. In the first work\, we
  present a Transfer Learning method to transfer the knowledge between two 
 domains (source and target) with disparate labels. Our approach exploits t
 he relationship between the source and the target labels to enhance the tr
 ansfer of the learned knowledge. We apply our methods to two NLP tasks: Ev
 ent Typing and Text Classification. In our second work\, we address the pr
 oblem of modeling the tasks with evolving type ontologies. We present a Ze
 ro-Shot Fine-Grained Entity Typing (ZS-FGET) approach that exploits the Wi
 kipedia description of the type to construct the representation of that ty
 pe. Then\, the type can be recognized requiring zero training examples. Si
 nce FGET deal with a large number of types organized into a hierarchy\, Di
 stant Supervision is employed to automatically collect training data\, lea
 ding to significant label noises. In our final work\, we focus on the hier
 archical nature of the fine grain entity types and propose an FGET framewo
 rk with a ranking-based weak supervision objective that ranks the types in
  the best path among the candidate paths higher than the incorrect types. 
 We further leverage additional type connections that are not presented in 
 the type hierarchy to improve the training and type inference and present 
 some preliminary results.\n\nMajor Advisor: Xiaoli Fern\nCommittee: Prasad
  Tadepalli\nCommittee: Stephen A Ramsey\nCommittee: Liang Huang\nGCR: Bret
 t Tyler
GEO:44.567164;-123.278692
LOCATION:Kelley Engineering Center\, 3114
SUMMARY:PhD Preliminary Oral Exam – Rasha Obeidat
URL;VALUE=URI:https://events.oregonstate.edu/event/phd_preliminary_oral_exa
 m_rasha_obeidat
CATEGORIES:Lecture or Presentation
END:VEVENT
END:VCALENDAR