INTELLIGENT SYSTEMS RESEARCH GROUP by Lois C. Boggess - - PDF document

▶
intelligent systems research group
SMART_READER_LITE
LIVE PREVIEW

INTELLIGENT SYSTEMS RESEARCH GROUP by Lois C. Boggess - - PDF document

Nov 10, 2023 311 likes •548 views

INTELLIGENT SYSTEMS RESEARCH GROUP INTELLIGENT SYSTEMS RESEARCH GROUP by Lois C. Boggess lboggess@cs.msstate.edu Julian E. Boggess gboggess@cs.msstate.edu Susan M. Bridges bridges@cs.msstate.edu Bradley D. Carter carter@cs.msstate.edu R.

slide-1
SLIDE 1

INTELLIGENT SYSTEMS RESEARCH GROUP

INTELLIGENT SYSTEMS RESEARCH GROUP

by Lois C. Boggess lboggess@cs.msstate.edu Julian E. Boggess gboggess@cs.msstate.edu Susan M. Bridges bridges@cs.msstate.edu Bradley D. Carter carter@cs.msstate.edu

  • R. Dwight Hare

rdh1@ra.msstate.edu Julia E. Hodges hodges@cs.msstate.edu Joe Picone picone@isip.msstate.edu Department of Computer Science, College of Eng. Institute for Signal and Information Processing, College of Eng. Instructional Technology Laboratory, ERC Curriculum and Instruction, College of Edu. Mississippi State University

ABSTRACT

We are a group with both mutual and complementary interests and strengths, in cognition, language, large bodies of data, multiple modes

  • f communication between computer and humans, machine learning

and adaptable systems. We’ve built systems in which the computer is seen as an aid to the human, rather than as the primary actor. Typically our goal is to achieve best possible performance when time constraints are sub-optimal, data are imperfect or incomplete, and there are multiple plausible ways for a system to proceed at any point in execution. These interests are supported by a core competency in a number of related information processing technologies including speech recognition, signal and image processing, natural language processing, machine learning, and expert database systems.

slide-2
SLIDE 2

INTELLIGENT SYSTEMS RESEARCH GROUP FEBRUARY 8, 1996 ISRG OVERVIEW PAGE 2 OF 22

Gene Boggess Lois Boggess Susan Bridges Julia Hodges

Computer Science

Dwight Hare

College of Education

Joe Picone

  • Elect. and Comp. Eng.

Brad Carter Don Trotter

ERC

Computer Aided Instruction Is Multidisciplinary By Nature

slide-3
SLIDE 3

INTELLIGENT SYSTEMS RESEARCH GROUP FEBRUARY 8, 1996 ISRG OVERVIEW PAGE 3 OF 22

Intelligent Systems Research Needs A Diverse Collection of Expertise and Experience

Brad Carter Director of Education, ERC

  • Software Engineering
  • Software Metrics
  • Instructional Technology

Gene Boggess Computer Science

  • Cognitive Science
  • Neural Networks
  • Genetic Algorithms

Susan Bridges Computer Science

  • Expert Systems
  • Explanation-Based Learning
  • Hybrid Systems

Julia Hodges Computer Science

  • Knowledge Bases
  • Database “Mining”
  • Machine Learning

Lois Boggess Computer Science

  • Natural Language
  • Very Large Corpora
  • Intelligent Tutoring

Dwight Hare Curriculum and Instruction

  • Learning and Pedagogy
  • Classification
  • Educational Policy

Joe Picone Computer Engineering

  • Speech Recognition
  • Statistical Modeling
  • Signal Processing
slide-4
SLIDE 4

INTELLIGENT SYSTEMS RESEARCH GROUP FEBRUARY 8, 1996 ISRG OVERVIEW PAGE 4 OF 22

An Intelligent Tutoring System from the Mid-80’s

Implementation:

❐ The Goal

Identify and provide remediation for a set of math topics at the sixth-grade level for adults preparing for the GED.

❐ The Cognitive Model

Extracted from State of Mississippi requirements for elementary teachers.

❐ Subjects

At least third grade competence in reading Some were students of a teacher instructing for GED Some were prisoners taking GED classes.

❐ Evaluation and Testing

System created pretest “on the fly” (filling slots in templates) so questions were not repeated. Students were given remediational material on subconcepts which appeared not to be mastered System created posttest.

❐ System “successful”

Statistically significant value added to student competence

slide-5
SLIDE 5

INTELLIGENT SYSTEMS RESEARCH GROUP FEBRUARY 8, 1996 ISRG OVERVIEW PAGE 5 OF 22

What did we learn? ❐ System did what it was designed to do ❐ No educational revolution followed ❐ A computerized book - with an attitude

slide-6
SLIDE 6

INTELLIGENT SYSTEMS RESEARCH GROUP FEBRUARY 8, 1996 ISRG OVERVIEW PAGE 6 OF 22

A Sampling of Mid-90’s Research

slide-7
SLIDE 7

INTELLIGENT SYSTEMS RESEARCH GROUP FEBRUARY 8, 1996 ISRG OVERVIEW PAGE 7 OF 22

AIMS: Automated Indexing at Mississippi State ❐ Unrestricted vocabulary ❐ Domain dependent ❐ Embeds human expertise ❐ Partners with human document analyst ❐ Tools to tune system to the way language is used within domain

slide-8
SLIDE 8

INTELLIGENT SYSTEMS RESEARCH GROUP FEBRUARY 8, 1996 ISRG OVERVIEW PAGE 8 OF 22

KUDZU: Knowledge Under Development from Zero Understanding ❐ Capture all the information present ❐ Open vocabulary ❐ Bootstrap a knowledge base Initially only metaknowledge of the domain Grows by reading the text

slide-9
SLIDE 9

INTELLIGENT SYSTEMS RESEARCH GROUP FEBRUARY 8, 1996 ISRG OVERVIEW PAGE 9 OF 22

Common Themes ❐ Learning the lingo: Understanding language

using cues in the language itself

❐ Information extraction:

from human experts from large bodies of data

❐ Interactive systems, human-centered interfaces, multiple modes of communication ❐ Data mining ❐ Machine learning - supervised/unsupervised ❐ Classification

slide-10
SLIDE 10

INTELLIGENT SYSTEMS RESEARCH GROUP FEBRUARY 8, 1996 ISRG OVERVIEW PAGE 10 OF 22

Characterization of Waste Assay Data ❐ Synthesis of data from multiple sources

  • database
  • multiple sensors
  • process history

❐ Determining confidence of characterization

  • consistency checking of data from multiple

sources

  • confidence associated with data sources

❐ Knowledge discovery

  • detecting patterns in data
  • learning classification rules based on patterns
slide-11
SLIDE 11

INTELLIGENT SYSTEMS RESEARCH GROUP FEBRUARY 8, 1996 ISRG OVERVIEW PAGE 11 OF 22

❐ One of the themes of “soft AI” is ❐ Genetic algorithms, neural networks: Data driven Good performance with good data Reasonable performance in the presence of incomplete or missing data, erroneous input Graceful degradation

slide-12
SLIDE 12

INTELLIGENT SYSTEMS RESEARCH GROUP FEBRUARY 8, 1996 ISRG OVERVIEW PAGE 12 OF 22

❐ Instead of systems which define to the world the boundaries within which the world must fit, we choose to build

  • Systems in which the domain boundaries are

fuzzy

  • Systems which do not impose limits in some

important aspects

  • Human-centered systems

Graceful accommodation to stretching the boundaries

slide-13
SLIDE 13

INTELLIGENT SYSTEMS RESEARCH GROUP FEBRUARY 8, 1996 ISRG OVERVIEW PAGE 13 OF 22

I S I P I S I P

s p ee c h s p ee c h

ISIP’s Focal Projects

  • An Integrated Services Transactions Processor That Supports

Advanced Telecommunications Interfaces such as an Asynchronous Transfer Mode (ATM) Digital Communications Link

Example: Telephone-Based Natural Language Query of Entertainment Archives Customer: “Give me all movies, uh, make that only the recent movies, directed by Martin Scorsese and starring Robert DeNiro, and oh, by the way, make that movies about gangsters only.” Computer: We have three titles available (the titles of the movies are shown on the television screen with real-time video of promo clips from each movie below the title). Please select a movie. Customer: “That one with the three guys looks good, I’ll take that one. I want it to start at 8:00 PM tomorrow.” Computer: (The promo clip for the selected movie starts playing on the television.) The movie titled GoodFellas starring Robert DeNiro and directed by Martin Scorsese will be delivered for viewing on your television on Thursday, September 25 starting at 8:00 PM. Thank you for using ISIP’s Entertainment

  • Server. Good-bye.

Local Central Office ATM (160 Mbps)

  • Voice
  • Video
  • Data (X Windows)

Unix Multiprocessor (Sparcstation 2000):

  • 8 Processors
  • 512 Mbytes of memory
  • videotape jukebox
slide-14
SLIDE 14

INTELLIGENT SYSTEMS RESEARCH GROUP FEBRUARY 8, 1996 ISRG OVERVIEW PAGE 14 OF 22

I S I P I S I P

s p ee c h s p ee c h Semi-Parser Language Model Tagged Text Natural Language Processing Request Generator Knowledge Extractor Filled Templates Netscape Requests Netscape Knowledge Extraction Flat Parsed Structures Speech Recognition Language Model Text Natural Language Understanding “Show me all the reports from the White House on Healthcare.”

slide-15
SLIDE 15

INTELLIGENT SYSTEMS RESEARCH GROUP FEBRUARY 8, 1996 ISRG OVERVIEW PAGE 15 OF 22

I S I P I S I P

s p ee c h s p ee c h

slide-16
SLIDE 16

INTELLIGENT SYSTEMS RESEARCH GROUP FEBRUARY 8, 1996 ISRG OVERVIEW PAGE 16 OF 22

I S I P I S I P

s p ee c h s p ee c h Sparcstation 1000

  • 8 processors
  • 500 Mbytes RAM
  • Dedicated Server

Demo Machine:

  • Sparcstation 5
  • ATM Interface
  • T1 Interface

Compute Server:

  • Sparc 20/60
  • 2 processors
  • 192 Mbytes RAM
  • 1 Gbyte local disk

The ISIP Computer Network: isip.msstate.edu

Outside World Sparcstation 5 Fileserver

  • 32 Mbytes RAM
  • 2 Gbyte local disk
  • 30 Gbytes external disk
  • 1 Exabyte Stacker (50 Gbyte)
  • 2 ethernets (Class C subnet)

B&W Printer

  • postscript
  • 600 dpi
  • 12 ppm

X/Audio Server:

  • Sparc SLC
  • 1 processor
  • 32 Mbytes RAM
  • 100 Mbyte disk
  • TCT DAT-Links
  • Network Audio

NCD X Terminals

  • 20” Color/Audio
  • 15” B&W

Color Scanner

  • 300 dpi
  • Adobe Photoshop

Sony DATs

  • 16-bit audio
  • Networked

28.8K Baud Dialup

  • LBW X Windows
  • 115K TTYs

URL: http://www.isip.msstate.edu FTP: ftp://ftp.isip.msstate.edu

slide-17
SLIDE 17

INTELLIGENT SYSTEMS RESEARCH GROUP FEBRUARY 8, 1996 ISRG OVERVIEW PAGE 17 OF 22

I S I P I S I P

s p ee c h s p ee c h

Basic Technology: A Pattern Recognition Paradigm Based On Hidden Markov Models

Search Algorithms:P Wt i Ot

( ) P Ot Wt

i

( )P Wt

i

( ) P Ot ( )

  • =

Pattern Matching: Wt i P Ot Ot 1 –

… Wt

i

, , ( ) , [ ]

Signal Model: P Ot

Wt

1 –

Wt Wt

1 +

, , ( ) ( )

Recognized Symbols: P S O

( ) max arg

T

P Wt

i

Ot Ot

1 –

… , , ( ) ( )

i

∏

=

Language Model: P Wt i

( )

Prediction

slide-18
SLIDE 18

n i: z

NP [case:obj]

def:+ form: acc num: pl type: bodypart

The doctor examined the patient’s knees. Orthographic:

d A k t exr dh ex I g z ae m I n d dhex pH e-I sh I nt s /# dh i # d A k t exr # I g z ae m ex n + d # dh i # p e sh ex n t + z # n i + z #/

Phonetic: Phonemic: VP ∃(X) & ∃(Y) & ∃(Z) & doctor(X) & patient(Y) & knees(Z) & part-of(Y,Z) & examined(X,Z) Logical:

S

NP [case:subj] NP N N Det

tense: past arg1: subj [type:caregiver] arg2: obj [type:bodypart]

V form: gen

num: sing type: patient

N form: nom

num: sing type: caregiver def:+

Det Lexical: Syntactic:

n i: z d A k t exr

1 2 0.5 1.5 5 kHz 0 kHz 4 kHz 3 kHz 2 kHz 1 kHz Time (secs)

dh ex I g z ae m I n d dhex pH e-I sh I nt s

Phonetic:

What can you do with all of this?

slide-19
SLIDE 19

INTELLIGENT SYSTEMS RESEARCH GROUP FEBRUARY 8, 1996 ISRG OVERVIEW PAGE 19 OF 22

I S I P I S I P

s p ee c h s p ee c h

Human Speech Recognition Performance Benchmarks On ARPA SLT CSR Corpora CSR’94: SPOKE 10 CSR’95: HUB-3

  • Overall human performance is at least an order of magnitude better

than machine performance Evaluation Group Vocabulary Open Closed Average 2.1 (0.7) 1.0 (0.6) Committee 1.2 (0.6) 0.5 (0.6) Evaluation Group Vocabulary Open Closed Average 2.2 2.1 Committee 0.5 0.5

slide-20
SLIDE 20

INTELLIGENT SYSTEMS RESEARCH GROUP FEBRUARY 8, 1996 ISRG OVERVIEW PAGE 20 OF 22

I S I P I S I P

s p ee c h s p ee c h

The JEIDA Japanese Common Speech Data Corpus

. Number of speakers 150 speakers 75 male speakers 75 female speakers Number of items per speaker monosyllables 178 isolated words 35 4-digit sequences 323 items Number of repetitions per item 4 repetitions of each item Range of speaker age 20 yrs. to 60 yrs. Amount of data 120 hours Number of Digital Audio Tapes 76 (120-minute tapes) Total number of utterances 193,800 utterances Number of channels/mic. type 2 (dynamic and condenser mics.) Anticipated size of final corpus (16-bit 16 kHz samples @ 1.0 secs per utterance) 6.5 Gbytes (13 CD-ROMs uncompressed)

slide-21
SLIDE 21

INTELLIGENT SYSTEMS RESEARCH GROUP FEBRUARY 8, 1996 ISRG OVERVIEW PAGE 21 OF 22

I S I P I S I P

s p ee c h s p ee c h

Automatic Generation of N-Best Proper Noun Pronunciations

  • • •
  • • •

u1,1 u2,1 u3,1 u2,1 u2,2 u3,2 u1,27 u2,26 u3,26 u3,27 u1,26 u1,1 u2,1 u3,1 u4,1 u5,1 u6,1 u7,1 u1,2 u2,2 u3,2 u4,2 u5,2 u6,2 u7,2 u1,26 u2,26 u3,26 u4,26 u5,26 u6,26 u7,26 u1,27 u2,27 u3,27 u5,27 u6,27 u7,27

  • • •
  • • •
  • • •
  • • •
  • • •
  • • •
  • • •

u1 u2 u414

  • • •
  • • •

u1 u2 u64

  • • •
  • • •
  • • •

u1,1 u2,1 u3,1 u2,1 u2,2 u3,2 u2,20 u3,20 u1,20

  • • •

u3,1 u3,2 u3,20

  • • •

Y S M N-Tuple Knowledge Variable Duration Context-Dependencies

{s: -voiced, ...} {m: +nasal,...} {(ay: +voiced,...), (ih: +voiced,...)} ...

Input # E H T Y M S

slide-22
SLIDE 22

INTELLIGENT SYSTEMS RESEARCH GROUP FEBRUARY 8, 1996 ISRG OVERVIEW PAGE 22 OF 22

I S I P I S I P

s p ee c h s p ee c h

What Differentiates ISIP Speech Research? ❐ Public Domain Software ❐ Extensive Web Archive ❐ Object-Oriented Signal Processing Software ❐ State-of-the-Art Performance Tasks ❐ Close Industrial Ties ❐ Next-Generation Statistical Models Based

  • n Chaotic Systems

Applicable to acoustic and language modeling Addresses a fundamental barrier in speech understanding