Skip to Content.
Sympa Menu

nl-uiuc - [nl-uiuc] [Fwd: [aivr] AIVRH Seminar 9/15 - Chilin Shih]

nl-uiuc AT lists.cs.illinois.edu

Subject: Natural language research announcements

List archive

[nl-uiuc] [Fwd: [aivr] AIVRH Seminar 9/15 - Chilin Shih]


Chronological Thread 
  • From: Margaret Fleck <mfleck AT cs.uiuc.edu>
  • To: nl-uiuc AT cs.uiuc.edu
  • Subject: [nl-uiuc] [Fwd: [aivr] AIVRH Seminar 9/15 - Chilin Shih]
  • Date: Mon, 12 Sep 2005 10:03:12 -0500
  • List-archive: <http://lists.cs.uiuc.edu/pipermail/nl-uiuc>
  • List-id: Natural language research announcements <nl-uiuc.cs.uiuc.edu>



-------- Original Message --------
Subject: [aivr] AIVRH Seminar 9/15 - Chilin Shih
Date: Fri, 9 Sep 2005 14:27:35 -0500
From: Kim Osmond
<kosmond AT cs.uiuc.edu>
To: 'AIVR'
<aivr AT cs.uiuc.edu>,
'Beckman AI faculty'
<ai-faculty AT scollins.ai.uiuc.edu>,
'Beckman AI students'
<ai-students AT scollins.ai.uiuc.edu>,
'cs-grads'
<cs-grads AT cs.uiuc.edu>
CC: 'Erna A Amerman'
<erna AT ad.uiuc.edu>



*AI/Vision/Robotics Seminar*



http://reason.cs.uiuc.edu/aivr



*Chilin** Shih***

*Department of EALC/Linguistics*

*UIUC*



Thursday, September 15, 2 p.m.

2405 Siebel Center



Prosody Learning and Generation



Words refer to *what* you say in speech. Prosody refer to *how* you say
it. Prosody is not written in text. Therefore, from the point of view
of building a text-to-speech system, prosody has to be generated from
scratch. This talk addresses one of the acoustic correlates of prosody,
the learning and generation of intonation (f0).



Many channels of information are carried in f0 contours, including
lexical information such as tones and accents, sentence intonation such
as question and statement, discourse functions such as focus and
information structure, as well as emotions such as happiness and anger.
How is all this information carried on a shared channel? How is it encoded?



Conflicts arise when all these pieces of information are represented
simultaneously. We show that speech prosody can be modeled as an
optimization problem, where the speaker attempts to minimize a
communication error term plus a term which behaves like the effort
required to communicate. This approach has yielded realistic models of
intonation in Mandarin, Cantonese, limited domain English dialogue, and
the imitation of distinctive rhetorical styles.





Kim Osmond

Department of Computer Science

University of Illinois

3316 Siebel Center

201 N. Goodwin Ave.

Urbana, Illinois 61801

217/244-4675



"The tragedy of life is not that it ends so soon, but that we wait so
long to begin it."

~ W.M. Lewis ~



_______________________________________________
aivr mailing list
aivr AT cs.uiuc.edu
http://lists.cs.uiuc.edu/mailman/listinfo/aivr



  • [nl-uiuc] [Fwd: [aivr] AIVRH Seminar 9/15 - Chilin Shih], Margaret Fleck, 09/12/2005

Archive powered by MHonArc 2.6.16.

Top of Page