BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//chikkutakku.com//RDFCal 1.0//EN
X-WR-CALDESC:GoogleカレンダーやiCalendar形式情報を共有シェ
 アしましょう。近所のイベントから全国のイベントま
 で今日のイベント検索やスケジュールを決めるならち
 っくたっく
X-WR-CALNAME:ちっくたっく
X-WR-TIMEZONE:UTC
BEGIN:VEVENT
SUMMARY:School of IT seminar slot (compulsory if scheduled)
DTSTART;VALUE=DATE-TIME:20260416T110000Z
DTEND;VALUE=DATE-TIME:20260416T115000Z
UID:328561060194
DESCRIPTION:Speaker: Francois Meyer\n\nTitle:\n\nData-Efficient Language M
 odelling for Low-Resource Languages\n\nAbstract:\n\nProgress in language m
 odelling has been driven by scaling data and\nmodel size\, but this approa
 ch is not feasible for most African\nlanguages. In this talk\, I will pres
 ent our work on developing\ndata-efficient language models - architectures
  and training algorithms\nthat improve performance on limited training dat
 a. I will present\nexamples of how linguistically informed modelling\, whi
 ch targets and\nleverages the linguistic properties of specific languages\
 , can improve\nsample efficiency. Finally\, I will discuss the emerging in
 tersection\nbetween low-resource NLP and developmentally inspired NLP\, ex
 ploring\nhow insights from human language learning can help us build more\
 nefficient models.\n\nBiography:\n\nFrancois Meyer is a Lecturer in the Co
 mputer Science Department at the\nUniversity of Cape Town and co-investiga
 tor in the UCT NLP research\ngroup. His research is on data-efficient lang
 uage modelling and\nlinguistically informed sub-word tokenisation. He comp
 leted his PhD at\nthe University of Cape Town and previously obtained a ma
 sters in AI at\nthe University of Amsterdam.
LOCATION:John Day LT1
END:VEVENT
END:VCALENDAR