Transcript PowerPoint Presentation - Language in Cognitive Science

Language in Cognitive Science
Research Areas for Language





Computational models of speech production
and perception
Signal processing for speech analysis,
enhancement and transformation applications
Algorithms and models for speech recognition
and synthesis
Speaker and language recognition
Natural language processing for speech
understanding and generation (cont.)
More Research Areas…



Information retrieval and summarization
Computational models of discourse and dialog
Evaluation of human and computer system
performance
--Research is currently being carried out by
practitioners of artificial intelligence,
computer science, electronic engineering,
linguistics, phonetics, and psychology.
Language Technology (LT)
Two Categories
1.
Spoken Language
2.
Written Language
4.
Two Constructs
Sending Language (Acquisition
5.
Receiving Language (Generation & Output)
3.
&Translation)
Spoken Language
Only in Spoken Language (not written):
-- filled pauses
-- false starts
-- hesitations
-- ungrammatical constructions
Spoken, cont.
Optical Information
-- facial expressions (eye contact,
movement of eyebrows, smiles or
frowns, etc.)
-- gestures (hand movement, body
positioning, posture, etc.)
LT terms



Prosody—acoustic structure that extends over
several segments or words (indicates user’s
intention)
-stress, intonation (tone), & rhythm
Modeling Dynamics—accounting for
differences in individual articulators
Multimodality—dealing with several principles
and aspects of speech/language
Approaches in LT--Templates

Use of Templates
-pre-programmed
-domain & task specific
-sets probabilities, restrictions,
allowances, distribution
Approaches in LT-Components

Use of components (an example)
In attempting to send language:
(1) Generate the message. This
would require a planning component
(brain forming ideas) and a syntatic
generation component (spellcheck)
(2)Output-requires some type of
interface
Approaches in LT—
Algorithms and Models


Use of Algorithms: essential to setting up
templates for language—are what enables
programmer to set up the restrictions,
distributions, etc…
Use of Models (an example)
Acoustic Model: Template design for
measuring the prosody of a given sample of
language (transition probabilities, output
distributions, etc..)
Some Tools for LT



CSLU Toolkit: involves speech recognition,
natural language understanding, speech
synthesis, facial animation technologies
Computerized Speech Labs (CSL): 2 models
(4300B & 4100) less sophisticated than CSLU
toolkit
Winpitch and Winpitch II: spectrographic
display, prosodic morphing