BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Putting Linguistics into Speech Recognition: The Regulus Grammar C
 ompiler - Manny Rayner\, Powerset.com/Geneva University
DTSTART:20061020T110000Z
DTEND:20061020T120000Z
UID:TALK5572@talks.cam.ac.uk
CONTACT:NLIP Seminars
DESCRIPTION:In this talk\, I will present an overview of Regulus\, an Open
  Source\nplatform that has been under development since 2001 at NASA Ames 
 Research\nCenter and Geneva University. Regulus contains tools and resourc
 es to\nsupport construction of grammar-based language models for\nmedium-v
 ocabulary spoken dialogue applications. Language models are\ncompiled out 
 of general linguistically motivated unification grammars\nusing example-ba
 sed methods driven by small corpora\, typically of a\nfew hundred examples
 . I will describe the compilation process\, present\nresults demonstrating
  that Regulus-derived language models are very\ncompetitive with statistic
 al ones derived from the same small training\nsets\, and show how the well
 -defined Regulus compilation process\nenables systematic investigation of 
 the properties of grammar-based\nrecognisers.\n\nRegulus has already been 
 used for several large projects\, including\nNASA's Clarissa astronaut ass
 istant (as far as we know\, so far the\nonly speech enabled system in spac
 e)\, and Geneva University's MedSLT\,\na multilingual medical speech trans
 lator. I will close by giving a\nshort demo of MedSLT.
LOCATION:SW01 Computer Laboratory
END:VEVENT
END:VCALENDAR