T. V. Raman

Cambr idge ResearchLab

Digital Equipment Corp.

Bldg 650, OneKendall Square

Cambr idge MA 02139

E-mail : [email protected]

Voice-mail : 1 (617) 692-7637

Fax : 1 (617) 692-6650

Ab stract Intro duction

Screen-reader s |computer software thatenable s a vi su-

A screen-reade r i s a computer application de s igned to

ally impaire d us er to read the contents of a vi sual

providespoken fee dbackto a vi sually impaire d us er.

di splay| have b een available for more than a decade.

Screen-readers have b een available s ince the mid-80's.

Screen-reader s are s eparate f rom theuserapplication.

Dur ingthe 80's, suchapplications relie d on thecharacter

Cons equently,they have little or no contextual inform-

repre s entation of the contentsofthe screen to pro duce

ation aboutthe contentsofthe di splay.Theauthor has

thespoken fee dback. The adventofbitmap di splays

us e d traditional screen-readingapplications for the last

le d to a complete breakdown of thi s approach, s ince the

veyears. Thede s ign of the sp eech-enablingapproach

contentsofthe screen were now lightanddark pixels.

de scr ib e d here has b een implemente d in Emacsp eak to

A s igni cant amount of re s earchanddevelopmenthas

overcomemanyoftheshortcomings hehas encountere d

b een carr ie d outtoovercomethi s problem and provide

with traditional screen-reade rs.

sp eech-acce ss tothe Graphical Us er Interf ace (GUI).

Theapproach used byEmacsp eak i s very di erent f rom

The best and p erhaps the most complete sp eech acce ss

that of traditional screen-reade rs . Screen-reade rs allow

system tothe GUI i s Screenreader /2 (ScreenReader For

theusertolisten tothe contentsapp ear ing in di erent

OS/2) developed by Dr. Jim Thatcher atthe IBM Wat-

partsofthe di splay; buttheuserisentirely re sp ons-

son Re s earch Center [Tha94]. Thi s package provides

ible for buildinga mental mo del of the vi sual di splay in

robust sp oken acce ss toapplications under the OS2

order tointerpret whatanapplication i s tryingto con-

Presentation Manager and Windows 3.1. Commercial

vey.Emacsp eak, on theother hand, do e s not sp eak the

package s for Microsoft Windows 3.1 providevarying

screen. Instead, applications provideboth vi sual and

levels of sp oken acce ss tothe GUI. TheMercator pro-

sp eech fee dback, andthe sp eech fee dbackisde s igned to

ject [ME92, WKES94, MW94, Myn94]has fo cus e d on

b e sucientbyits elf.

providingspoken acce ss tothe X-Windows system.

Thi s approachreduce s cognitive load on theuserand

A common feature of traditional DOS-bas e d screen-

i s relevantto providing general sp oken acce ss to in-

readers and sp eech acce ss package s tothe GUI i s their

formation. Pro ducingspoken output f rom within the

attempt to convey the contentsofthe vi sual di splay via

application, rather than sp eakingthe vi sually di splayed

sp eech. In f act, a s igni cant amountofthedevelopment

information, vastly improves the qualityofthespoken

e ort require d tode s ign sp eech-acce ss package s tothe

fee dback. Thus, an application can display its re sults

GUI has concentrate d on building up robust o -screen

in a vi sually pleas ingmanner; the speech-enabling com-

models |a data structure that repre s entsthe contents

ponent renders the same in an aurally pleas ingway.

of the GUI's vi sual di splay. Construction of such an o -

screen mo del helps screen-reade rs regain the groundthey

lost due tothe advent of graphical di splays. However,

Keywords: Sp eechInterf ace, Direct Acce ss, Sp oken

thenature of sp oken fee dback provided does not change.

Fee dback, Audio Formatting, Sp eech as a rst-class I/O