BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Confidence estimation for attention-based encoder-decoder models f
 or speech recognition - Qiujia Li (University of Cambridge)
DTSTART:20220606T110000Z
DTEND:20220606T120000Z
UID:TALK175244@talks.cam.ac.uk
CONTACT:Dr Jie Pu
DESCRIPTION:*Abstract*: Confidence scores have been an intrinsic part of a
  conventional speech recogniser. As end-to-end ASR models such as attentio
 n-based encoder-decoder models become increasingly popular\, it is of grea
 t interest to develop reliable confidence estimators for various downstrea
 m tasks. In this talk\, I will present the confidence estimation module (C
 EM) for token/word-level confidence scores\, and the residual energy-based
  model (R-EBM) for utterance-level confidence scores for attention-based m
 odels. Interestingly\, R-EBM can also help improve the ASR performance. Fu
 rthermore\, some effective techniques for generalising these model-based c
 onfidence estimators to out-of-domain data will be discussed.\n\n*Bio*: Qi
 ujia Li is a fourth-year PhD student at the University of Cambridge\, advi
 sed by Prof. Phil Woodland. He obtained his BA and MEng also from Cambridg
 e University. His research interests lie primarily in speech processing an
 d machine learning\, including end-to-end speech recognition\, confidence 
 estimation and speaker diarization. He has published more than a dozen pap
 ers at ICASSP\, Interspeech\, SLT\, ASRU\, NeurIPS and ICCV\, of which two
  won the best student paper awards at ASRU 2019 and SLT 2021. He previousl
 y worked as a research intern with Microsoft in 2018 and Google in 2020.
LOCATION:Zoom: https://eng-cam.zoom.us/j/81927138251?pwd=TVd3MXliV003dUdYV
 lFwU2NDWGpmdz09
END:VEVENT
END:VCALENDAR