VoiceCode Framework

1
VoiceCode Framework

• Alain Désilets
• Institute for Information Technology
• National Research Council of Canada
• Canada
• alain.desilets _at_iit.nrc.ca

2
What is VoiceCode?

• NRC lead Open Source initiative
• Goal
• to make programming-by-voice as easy and
  productive as with traditional input devices
• By
• developing a toolbox of synergistic, tested, and
  documented components that support current
  programming-by-voice best practices for all major
  programming environments and languages

3
Why should we do this?

• Programming by Voice (PBV) with current tools is
  like gargling with broken glass
• Eric Johansson
• 681 000 programmers in US alone ()
• 34 000 of which have RSI ()
• 3 500 increase annually
• () US Bureau of Labor Statistics
• () Based on 5 incidence

4
Why go Open Source?

• Little incentive for commercial development or
  development by a single organisation
• BUT
• Pool of users/co-developpers
• Scratch your own itch
• Existing body of public domain code

5
PBV Today Personal bag of tricks

• large startup cost (6 mo to 1 yr)
• scavenge-hunt multiple mailing lists, archives,
  etc
• no well document wisdom
• independently developed bits and pieces
• environment and/or language specific
• no standard
• little synergy between components

6
PBV Tomorrow Standard Toolbox

• small startup cost (1 mo)
• single stop shopping
• well documented wisdom
• standard toolbox
• supports all major environments and languages
• high synergy between components

7
Project Status

• Nov 99
• VoiceCode Open Source project proposed on
  VoiceCoder mailing list
• March 2000
• 9 Core members
• 30 Attendees to 1st VoiceCode design meeting
• Agreement on best practices to be supported
• Plan for implementation
• Piecemeal implementation to starts in April 2000

8
We need a plan

• Goals for todays workshop
• Achieve consensus on
• architecture framework
• list of components to be implemented
• supported languages
• supported programming environments
• priorities
• Who does what?

9
HCI principles

• Do what I mean, not what I say
• Flexible wording
• Continuous speech, pause wherever

10
S/W engineering principles

• SR vendor independence
• OS independence
• Cross environment
• Cross language
• Customisability

11
Architecture

• Context Sensitive Magic Words used to bind
  everything together.
• Allows user to
• enter template code
• navigate code
• invoke other utilities
• by speaking in a continuous and natural way

12
Architecture
Programming Environment
SR System
Macros
Continuous Speech
Recognised Utterance

• SAPI
• Keyboard events

Magic Word Context Info
Actions
CSMW Engine
13
Architecture

• CSMW engine to
• avoid reinventing the wheel for each programming
  environment/language
• ensures consistency of wording across those
• cross-fertilisation between users of different
  environments
• but implementation of actions is specific to
  each environment

14
Development framework
15
Components

• CSMWs
• Infrastructure for interpretation
• CSMWs for template code in major languages
• CSMWs for navigation in major languages
• CSMWs for invoking other utilities
• Correction infrastructure
• SR Training utilities

16
Components

• Other utilities
• Cache Pad
• VG style symbol translation
• Breadcrumbs
• Navigate by line number
• Continuous scrolling
• etc...

17
Components

• S/W Engineering stuff
• Customisation tool
• Automated regression testing
• Documentation

18
Agenda

• Architecture and development framework
• List of components
• Priorities
• Who does what?
• Financial support