BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Long Sequence-to-Sequence Summarization: Efficient Transformer Mod
 els &amp\; Complementary Techniques - Potsawee Manakul\, Cambridge Univers
 ity Engineering Department
DTSTART:20210525T110000Z
DTEND:20210525T120000Z
UID:TALK157531@talks.cam.ac.uk
CONTACT:Dr Kate Knill
DESCRIPTION:*Abstract*: Transformer-based models have achieved state-of-th
 e-art results in a wide range of tasks including document summarization. T
 ypically these systems are trained by fine-tuning a large pre-trained mode
 l to the target task. One issue with these transformer-based models is tha
 t they do not scale well in terms of memory and compute requirements as th
 e input length grows. Thus\, for long document summarization\, it can be c
 hallenging to train or fine-tune these models. In this talk\, first\, we w
 ill cover some recent efficient transformer models for sequence-to-sequenc
 e tasks\, including the motivation behind the design choice as well as the
 ir performance on summarization tasks. Second\, the talk will cover altern
 ative techniques that are complementary to efficient architectures. The ta
 lk will discuss CUED systems that were successful in the Spotify Podcast S
 ummarization Challenge 2020.\n\n*Bio*: Potsawee Manakul is a 2nd-year PhD 
 student supervised by Prof. Mark Gales\, in the Speech Group\, Department 
 of Engineering\, University of Cambridge. His primary research interests i
 nclude text summarization\, summary assessment\, and broader natural and s
 poken language processing. He obtained B.A. and M.Eng. degrees from the Un
 iverisity of Cambridge\, where he studied information and computer enginee
 ring.\n\nRelated papers:\n\n[1] "Long-Span Dependencies in Transformer-bas
 ed Summarization Systems" (ACL2021)\, Link: https://arxiv.org/abs/2105.038
 01\n\n[2] "CUED_speech at TREC 2020 Podcast Summarisation Track"\, Link: h
 ttps://arxiv.org/abs/2012.02535\n
LOCATION:Zoom: https://zoom.us/j/95352633552?pwd=RzJVK2UzOGZyNU5mVHd1Y1VPT
 2tDUT09
END:VEVENT
END:VCALENDAR