BEGIN:VCALENDAR
CALSCALE:GREGORIAN
VERSION:2.0
METHOD:PUBLISH
PRODID:-//Drupal iCal API//EN
X-WR-TIMEZONE:America/New_York
BEGIN:VTIMEZONE
TZID:America/New_York
BEGIN:DAYLIGHT
TZOFFSETFROM:-0500
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=2SU
DTSTART:20070311T020000
TZNAME:EDT
TZOFFSETTO:-0400
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:-0400
RRULE:FREQ=YEARLY;BYMONTH=11;BYDAY=1SU
DTSTART:20071104T020000
TZNAME:EST
TZOFFSETTO:-0500
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
SEQUENCE:1
X-APPLE-TRAVEL-ADVISORY-BEHAVIOR:AUTOMATIC
UID:204761
DTSTAMP:20250220T145002Z
DTSTART;TZID=America/New_York:20250305T100000
DTEND;TZID=America/New_York:20250305T110000
URL;TYPE=URI:https://www.wpi.edu/news/calendar/events/computer-science-depa
 rtment-phd-defense-yichuan-li-text-centric-representation-learning-integra
 ting
SUMMARY:Computer Science Department, PhD Defense , Yichuan Li " Text-Centri
 c Representation Learning: Integrating Multimodal Knowledge to Overcome La
 beled Data Scarcity  "
DESCRIPTION:\n\nYichuan Li\nPhD Candidate\nWPI – Computer Science Departmen
 t\n\nWednesday, March 5, 2025\nTime: 10:00 a.m. – 12:00 p.m.\nLocation: Go
 rdon Library , Room 303\nZoom Link:https://wpi.zoom.us/j/4166606557?omn=96
 360839200\nDissertation Committee:\nDr. Kyumin Lee, PhD Advisor, WPI – Com
 puter Science\nDr. Nima Kordzadeh, WPI – Business School\nDr. Xiangnan Kon
 g, WPI – Computer Science\nDr. Xiaozhong Liu, WPI – Computer Science\nDr. 
 Kaize Ding, Assistant Professor, Northeastern University\n\nAbstract:\nAdv
 ancements in textual representation learning often face the challenge of r
 equiring extensive labeled data, which is resource-intensive and impractic
 al in dynamic fields like fake news detection and social media analytics. 
 This dissertation introduces a paradigm shift from traditional "text-only"
  approaches to a "text-centric" methodology, integrating auxiliary knowled
 ge from diverse sources and modalities to enhance text representation with
 out relying heavily on labeled data. The first contribution emphasizes inc
 orporating human prior knowledge and domain expertise. A novel framework e
 mploys multi-source domain adaptation and weak supervision to detect fake 
 news early, transferring knowledge from well-labeled source domains to tar
 get domains with limited labeled data.\nAdditionally, meta-learning and co
 ntrastive learning techniques are utilized to reduce noise in augmented da
 ta, improving text classification by reweighting and refining feature repr
 esentations. The second contribution explores the fusion of numerical feat
 ures with textual data in causal analyses of fake news dissemination on so
 cial media. A causal inference model combining textual and numerical covar
 iates identifies key lexicons and posts that drive misinformation spread, 
 informing more effective intervention strategies. The final contribution p
 resents GRENADE, a groundbreaking graph-centric language model designed fo
 r self-supervised representation learning on text-attributed graphs. By al
 igning pre-trained language models with graph neural networks, GRENADE cap
 tures both textual semantics and structural context, demonstrating efficac
 y in learning expressive, generalized representations in citation and prod
 uct networks.\n
END:VEVENT
END:VCALENDAR