BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Recognition and Correction of Voice Web Search Queries - Keith Ver
 tanen (University of Cambridge)
DTSTART:20091005T100000Z
DTEND:20091005T110000Z
UID:TALK19956@talks.cam.ac.uk
CONTACT:Emli-Mari Nel
DESCRIPTION:In this work we investigate how to recognize and correct voice
  web search queries.  We describe our corpus of web search queries and sho
 w how it was used to improve recognition accuracy.  We show that using a s
 earch-specific vocabulary with automatically generated pronunciations is s
 uperior to using a vocabulary limited to a fixed pronunciation dictionary.
   We conducted a formative user study to investigate recognition and corre
 ction aspects of voice search in a mobile context.  In the user study\, we
  found that despite a word error rate of 48%\, users were able to speak an
 d correct search queries in about 18 seconds.  Users did this while walkin
 g around using a mobile touch-screen device.\n\nFor more details\, there i
 s a "paper and video":http://www.keithv.com/pub/webvoice/\n\nAlso see chap
 ter 6 of my "thesis":http://www.keithv.com/pub/phd/\n\nThis will be a shor
 t talk with plently of time for questions.  I will demo the system after t
 he talk.\n
LOCATION:TCM Seminar Room\, Cavendish Laboratory\, Department of Physics
END:VEVENT
END:VCALENDAR