BEGIN:VCALENDAR
PRODID:-//eluceo/ical//2.0/EN
VERSION:2.0
CALSCALE:GREGORIAN
BEGIN:VEVENT
UID:www.tcs.tifr.res.in/event/535
DTSTAMP:20230914T125928Z
SUMMARY:Discriminative Pronunciation Modeling for Dialectal Speech Recognit
 ion
DESCRIPTION:Speaker: Izhak Shafran\n​ (Google Inc.\nUnited States of Amer
 ica)\n\nAbstract: \nAbstract: Speech recognizers are typically trained wit
 h data from a standard dialect and do not generalize to non-standard diale
 cts. Mismatch mainly occurs in the acoustic realization of words\, which i
 s represented by acoustic models and pronunciation lexicon. Standard techn
 iques for addressing this mismatch are generative in nature and include ac
 oustic model adaptation and expansion of lexicon with pronunciation varian
 ts\, both of which have limited effectiveness. We present a discriminative
  pronunciation model whose parameters are learned jointly with parameters 
 from the language models. We tease apart the gains from modeling the trans
 itions of canonical phones\, the transduction from surface to canonical ph
 ones\, and the language model. We report experiments on African American V
 ernacular English (AAVE) using NPR's StoryCorps corpus. Our models improve
  the performance over the baseline by about 2.1% on AAVE\, of which 0.6% c
 an be attributed to the pronunciation model. The model learns the most rel
 evant phonetic transformations for AAVE speech.\n\nBio: Izhak Shafran is a
  speech researcher\, who has been working on acoustic modeling and large v
 ocabulary speech recognition since 1996. Before joining Google\, he was an
  Associate Professor and a member of the Center for Spoken Language Proces
 sing at OHSU\, where his focus was on medical application specifically on 
 Parkinsons' disease\, depression and mild congitive impairment. He graduat
 ed from University of Washington in Seattle in 2001 and subsequently worke
 d at AT\\&T Research Labs at Florham Park with the speech algorithms group
 . In summer of 2006\, he was a visiting professor at Univeristy of Paris-S
 outh\, working at LIMSI. Subsequently\, he was a research faculty at the C
 enter for Language and Speech Processing (CLSP) in Johns Hopkins Universit
 y. He received an NIH Career Development Award in 2010. He started his res
 earch career at Tata Institute of Fundamental Research in Radio Astronomy.
 \n
URL:https://www.tcs.tifr.res.in/web/events/535
DTSTART;TZID=Asia/Kolkata:20140903T150000
DTEND;TZID=Asia/Kolkata:20140903T160000
LOCATION:AG-80
END:VEVENT
END:VCALENDAR