BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Typological Feature Prediction and Blinding for Cross-Lingual NLP 
 - Johannes Bjerva (Aalborg University)
DTSTART:20210521T110000Z
DTEND:20210521T120000Z
UID:TALK160303@talks.cam.ac.uk
CONTACT:Huiyuan Xie
DESCRIPTION:Join Zoom Meeting\nhttps://cl-cam-ac-uk.zoom.us/j/99610057235?
 pwd=dGV0bENBa1lxTmFaa28rSmg2dFlldz09\n\nMeeting ID: 996 1005 7235\nPasscod
 e: 525413\n\nI will discuss the usefulness of typological features in NLP\
 , particularly in cross-lingual settings. On the one hand\, typological fe
 atures from databases such as the World Atlas of Language Structures (WALS
 ) seem promising for cross-lingual NLP\, as annotations for useful aspects
  of language exist even for very low-resource languages. Furthermore\, mis
 sing features in WALS can be predicted with relatively high success\, and 
 has been the focus of much recent work (e.g. in the SIGTYP 2020 shared tas
 k). When it comes to application of these features\, however\, previous wo
 rk has only found minor benefits from using typological information in act
 ual NLP modelling. In recent work (to appear at EACL 2021)\, we hypothesis
 ed that these minor gains might stem from that a model trained in a cross-
 lingual setting picks up on typological cues from the input data\, thus ov
 ershadowing the utility of explicitly using such features. We verify this 
 hypothesis by blinding a model to typological information\, and investigat
 e how cross-lingual sharing and performance is impacted. While this sheds 
 some light on the matter\, I want to further explore the question of the u
 sefulness of typological feature prediction in general\, and the use of su
 ch features in NLP.
LOCATION:Virtual (Zoom)
END:VEVENT
END:VCALENDAR