UCLA Academic Technology Services

ATS HomeATS ServicesATS ClassesATS ContactsATS Jobs
ATS Search

 

 

Text of Slides From Monsters In The Portal

 

Monsters in The Portal

Patrick Burke & Kevin Price

Disabilities and Computing Program, UCLA

Los Angeles, CA


Traditional Interaction With The Computer

  • Mouse

  • Keyboard

  • Monitor

Future Of Computer Interaction

  • The traditional form of computer interaction is impossible for some, difficult for others, and not the preference for many.

  • Many people would like to interact with computers as represented in science fiction.

  • Bill Gates: "Speech is not just the future of Windows, but the future of computing itself."

3 Things Need To Be In Place

  • 1. Accurate speech recognition technology that is easy to use.
  • 2. Text to Speech technology that is easy to understand.
  • 3. Smooth integration of the technologies.

What Exists Today?

  • The science fiction model of interaction with a computer without a mouse, monitor, or keyboard is not reality yet
  • Tools are now available that allow users to interact with computers both hands-free and eyes-free.

Monster Tools That Work Together

  • Dragon NaturallySpeaking Professional Speech Recognition Technology
  • JAWS (Job Access with Speech) Speech Output technology
  • JawBone software program - integrates Dragon and JAWS

Speech Recognition Concepts

  • Continuous vs. Discreet Speech
  • Large Vocabulary vs. Limited Vocabulary
  • Speaker Independent vs. Speaker Dependent
  • Global vs. Specific
  • Command & Control vs. Dictation
  • Matching vs. Natural Language Understanding

How Speech Recognition Technology Works

  • Analog converted to digital
  • Digital converted by computer software to text characters
  • - sophisticated pattern matching
  • - best guess algorithm
  • Breaks sound into phonemes and pieces them back together
  • Grammatical Modeling

    • What Speech Files Are Made Of

  • Acoustic model (pronunciation of phonemes trained by individual user)
  • Language Model (mathematical probability)
  • Active Vocabulary
  • Backup Dictionary
  • Constrained Grammar (looks first here)

    • Improving Speech Recognition

  • Completing individualized training
  • Adding vocabulary used frequently
  • Correcting mistakes during dictation
  • Creating macros
  • Purchasing specialized versions

    • Improving Speech Recognition (Continued)

  • Using good techniques.
  • - enunciating (but not over enunciating)
  • - proper positioning (microphone and body)
  • - resting vocal cords, drinking liquids
  • - avoiding areas with loud sharp noises
  • Using technology that meets optimal requirements.
  • Using the system consistently so commands are learned. More hands-free, more commands needed.

    • Dragon NaturallySpeaking Characteristics

  • Continuous, large-vocabulary speech recognition program introduced in 1997.
  • 160,000-word active vocabulary out of a 250,000-word dictionary.
  • 150 words per minute dictation is possible.
  • 98% -99% accuracy reported possible.
  • Uses enunciated, continuous speech, not conversational speech.
  • Uses Microsoft's Active Accessibility.

    • What Is Required To Use NS?

  • Minimum Technical Requirements
  • -350 MHz PC
  • -128 Megabytes of RAM
  • -Good sound card (Turtle Beach, Creative)
  • -Primary, Secondary Cache improves performance
  • -Noise Canceling Microphone

Other Requirements of NS

  • Fluid speech (for continuous speech products)
  • Basic knowledge of computers
  • Fairly high cognitive ability
  • Patience

JAWS Characteristics

  • Most widely used screen-reading software
  • Introduced in 1995
  • Development in Windows based on JAWS For DOS
  • Speech output quality dependent upon speech synthesizer

JAWS Characteristics

  • No configuration needed to read applications that use standard Windows controls and conventions.
  • Capable of high degree of configuration for many non-standard applications.
  • High degree of user customization available.

What Is Required To Use JAWS?

  • A computer of average power (more processor and memory resources needed for software-based speech output).
  • Good keyboarding skills (including multiple-key combination commands).
  • Lengthy time investment for beginning users.
  • Basic knowledge of computers.
  • Fairly high cognitive ability.

JawBone Characteristics

  • Optimized combination of JAWS and Dragon configuration files allowing the programs to function simultaneously (introduced 1999).
  • Minimizes command conflicts between JAWS and NS.

JawBone Characteristics

  • Includes voice macros to carry out JAWS commands and some keystrokes to carry out Dragon operations.
  • Works optimally in Dragon dictation window (limited functionality and reduced recognition accuracy in other applications).
  • Most efficient if some keyboard control is available.

What Is Required To Use JawBone?

  • Powerful computer
  • Fluid speech and knowledge of macro command names
  • Understanding of fundamentals of both JAWS and NS
  • Fairly high cognitive ability
  • Keyboarding skill useful but not necessary
  • Patience

Problems With Monster Setup

  • Privacy issues
  • More hands-free and eyes-free, the higher the learning curve
  • Wear and tear on vocal cords
  • Global use not possible yet
  • Expectations higher than reality
  • Need high-end computer and specialized software (expensive)

Advantages Of Monster Setup

  • Technologies are getting better and may be integrated into future devices.
  • Price is coming down. Not $10,000 - $15,000 but around $5,000.
  • Offers hands-free and eyes-free options for specialized work environments.
  • Helps those who have verbally oriented writing styles.
  • It is a good option for people with visual impairments or multiple disabilities.