BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Incorporating Structure into NLP Models with Graph Neural Networks
  -  Michael Schlichtkrull (University of Cambridge)
DTSTART:20210514T110000Z
DTEND:20210514T120000Z
UID:TALK160300@talks.cam.ac.uk
CONTACT:Huiyuan Xie
DESCRIPTION:Many of the most interesting NLP applications require modellin
 g various structured sources in addition to text. In this talk\, I will di
 scuss how such structured data can be incorporated into neural NLP models 
 with graph neural networks. In the first part\, I will give a brief introd
 uction to the subject\, and talk about our results on modelling knowledge 
 bases for link prediction and question answering. I will also discuss the 
 correspondence to transformers\, along with some recent results on modelli
 ng tables for fact verification. The second part of my talk will be about 
 interpreting what these models learn. Graph neural networks are complex an
 d highly nonlinear models. They can help models benefit from structure\, b
 ut it can be difficult to understand which structures are useful\, how exa
 ctly the model uses the structure\, and why specific decisions are made. I
  will talk about our recent work on GraphMask\, an interpretability techni
 que for graph neural networks. GraphMask produces rationales explaining wh
 ich parts of a graph a given model relies on\, both for individual example
 s and on the dataset level.\n\nJoin Zoom Meeting\nhttps://cl-cam-ac-uk.zoo
 m.us/j/91409349297?pwd=NUVVOW5SUTNpZ2w1UkJicmhLY3MzZz09\n\nMeeting ID: 914
  0934 9297\nPasscode: 612874
LOCATION:Virtual (Zoom)
END:VEVENT
END:VCALENDAR