A FLEXIBLE INTERFACE ARCHITECTURE FOR DIGITAL

TALKING BOOKS

Carlos Duarte, Lu

ıs Carric¸o, Hugo Sim

oes

LaSIGE, Department of Informatics, Faculty of Sciences of the University of Lisbon

Campo Grande, Edif

ıcio C6, 1749-016 Lisbon, Portugal

Keywords:

Adaptive Interfaces, Multimodal Interaction, Model-based Architecture, Digital Talking Books.

Abstract:

Talking books, besides helping the blind and print-disabled communities to have easier access to books, also

allow reading in situations when the vision becomes temporarily unavailable, for instance, while driving. In

this paper, we present a user interface model-based architecture for digital talking books. The books allow for

multimodal interaction in an effort to broaden the usage scope. Considering the execution platform, user and

usage characteristics, a great number of conﬁgurations for the books are possible. For the situations where the

most suited conﬁguration can not be forecast, an adaptable book can be the solution.

1 INTRODUCTION

Bringing together written and spoken words opens up

new ways of exploring books. A Digital Talking Book

(DTB) accomplishes it by combining the visual re-

production of the book with the audio reproduction

of the narration. In comparison to audio tapes with

analogue recordings of the books, the digital format

brings more than an improvement in the reproduc-

tion quality. New interaction possibilities, that be-

fore were unavailable or cumbersome, become real-

ity. Searching for a word no longer forces the user to

listen to (possibly) the whole book. Annotating and

navigating the book is also now made possible.

Besides the visually impaired persons, who are the

main target audience for DTBs, several other popu-

lation segments can beneﬁt from a platform such as

this. The multimodal interaction capabilities of the

book broaden the reading opportunities to situations

where the reader is engaged in visual occupying activ-

ities, such as driving or surveillance. Balancing DTB

modes and media can be explored to overcome the

cognitive limitations of human perception and atten-

tion (Gazzaniga et al., 1998).

The expansion of the traditional reading experi-

ence, reshaping it into an “immersion” in a multi-

media environment, can be envisioned within a mul-

timodal playback environment. The possibility of

creative combination of presentation elements, tak-

ing advantage of available media resources, offers the

support to new ways of telling stories and improving

learning (Carric¸o et al., 2003). The evolution is based

on the introduction of new multimedia elements, in

a coherent way, during “reading”. Possible enrich-

ments are the introduction of background music, en-

vironmental sounds related to the “scene of the ac-

tion”, images or videos to complement information

presented in the original source, and many more.

The availability of multimodal interaction on the

playback platform is desirable to complement this

multimedia presentation. In order to improve the cov-

erage, reliability and usability of the interface, it is

important to automatically adapt to user, task and en-

vironment parameters (Oviatt et al., 2004). A book’s

presentation and interaction should adapt to the users’

characteristics and preferences, to the environmental

conditions, and to the playback platform and interac-

tion devices available.

A framework for automatic production of DTBs,

that is ﬂexible enough to build books tailored to spe-

ciﬁc users and usage situations, has already been de-

veloped (Carric¸o et al., 2004). Nevertheless, in situa-

tions where the book designer is unable to predict the

characteristics of the users, the playback platform and

environment, the book must be able to adapt automat-

ically to those changing conditions. That is why we

need an adaptive DTB (Duarte and Carric¸o, 2004).

In the next section we present a brief overview of

related work, covering non-visual interfaces, adaptive

interfaces and DTB standards and navigation features.

146

Duarte C., Carriço L. and Simões H. (2004).

A FLEXIBLE INTERFACE ARCHITECTURE FOR DIGITAL TALKING BOOKS.

In Proceedings of the Sixth International Conference on Enterprise Information Systems, pages 146-151

DOI: 10.5220/0002636101460151

 SciTePress

We then identify some adaptation dimensions speciﬁc

to DTBs. We follow by describing the model-based

architecture of our user interface development frame-

work, relate it to the run-time execution of the inter-

face, and how it can be used to adapt to platform, user

and environment changes. We ﬁnish by drawing some

conclusions.

2 RELATED WORK

2.1 Digital Talking Books

DTBs are intended to give an easier access to books

to the blind and print-disabled community. Members

of those communities cooperated with several organi-

zations that developed DTB related standards. In Eu-

rope, the Daisy Consortium, with collaboration from

the European Blind Union developed one of those

standards. In the USA a similar work was conducted

by the National Information Standards Organization

(NISO) in collaboration with The National Library

Service for the Blind and Physically Handicapped.

From the cooperation between the Daisy Consortium

and the NISO resulted the most important DTB speci-

ﬁcation, the ANSI/NISO z39.86 (ANSI/NISO, 2002).

According to the NISO Document Navigation Fea-

tures List (NISO, 1999), a DTB should provide ba-

sic navigation capabilities (advancing one character,

word, line, sentence, paragraph or page at a time, and

navigation to speciﬁc segments of the DTB), fast for-

ward and reverse, reading at variable speeds, naviga-

tion through table of contents or a navigation control

ﬁle (allowing the user to obtain an overview of the

material in the book), reading notes, cross-reference

access, bookmarking, searching and others.

However, and wisely, no speciﬁc implementation

solutions are present in the standard. The solutions

must consider aspects related to the proposed speciﬁ-

cation, but also the non-visual nature of the targeted

environment.

2.2 Speech Interfaces

The work on non-visual interfaces can provide us with

clues on how to tackle some of the problems faced.

Voice browsers are devices that exhibit at least one

of the following characteristics: (1) can render web

pages in audio format; (2) can interpret speech for

navigation. Voice browsers and DTB interfaces share

some common problems:

• The audio format is a temporal medium. A visually

presented page can render simultaneously images,

tables and text, in a spatial format, which is quickly

processed by the perceptual human system. Spoken

text, however, can present only one word at a time.

• Issuing voice commands, and audio processing,

are activities that consume working and short-term

memory, conﬂicting with planning and problem

solving tasks. Visual information is processed by

separate cognitive systems (Christian et al., 2000).

• The unavoidable recognition errors.

However, the research in the multimodal systems

ﬁeld have made it clear that speech input is advan-

tageous under certain circumstances (Oviatt et al.,

2000). Studies (Van Buskirk and LaLomia, 1995;

Christian et al., 2000) point out that “the best tasks

for speech input were tasks in which the user has to

issue brief commands using a small vocabulary”.

The interaction characteristics of a DTB are advan-

tageous for the adoption of a speech interface: a rela-

tively small number of commands can be used to im-

plement the needed functionalities. However, some

limitations may arise, if, for instance, to follow a ta-

ble of contents entry, the user is forced to speak the

chapter’s title.

Research on the efﬁciency of speech as an input

mode is not conclusive, although showing an increase

in task completion time (Van Buskirk and LaLomia,

1995; Christian et al., 2000). Some of the recommen-

dations made for constructing voice browsers can be

adopted for the design of DTBs: links should be eas-

ily spoken text; links should be short (a few words);

avoid links with similar sounds; and develop alterna-

tives to numbered links, as these cause cognitive over-

load.

2.3 Adaptive Interfaces

An adaptive interface has been deﬁned as “a software

artefact that improves its ability to interact with a user

by constructing a user model based on partial experi-

ence with that user” (Langley, 1999). This means that

an adaptive interface must have generalization abili-

ties (because the adaptation is based only on partial

experience with past user interactions) and that the

adaptation is based on a user model. However we may

argue that relying only on the user model, as the ba-

sis for the adaptation may be insufﬁcient. There are

situations where the drive of the adaptation shouldn’t

originate from the user, but from external events or

environmental changes.

Several reviews of user models exist (Kok, 1991;

Kobsa, 2001), but we will focus only on two user

models here. A number of characteristics could be

used to identify users of a certain subgroup, called

a stereotype (Rich, 1989). Once the type is known,

the interface can be adapted to accommodate the user.

This approach can be used when producing a DTB

tailored to groups of users. Overlay models (Carr and

Goldstein, 1977) can be used to represent the users

knowledge (or preferences) on some domain. For

A FLEXIBLE INTERFACE ARCHITECTURE FOR DIGITAL TALKING BOOKS

147

this purpose a set of concept-value pairs is needed.

The concepts form elementary pieces of knowledge

for the given domain, and the value associated with it

represent how well the concept is known to the user.

We can envision the use of overlay models whenever

producing a DTB that could be tailored to individual

users.

Adaptive interface systems cover many areas, with

educational applications and on-line information sys-

tems being the most popular (Brusilowsky, 2001). We

can relate the creation of adaptive DTBs to both of

these areas, very clearly when producing a DTB with

educational purposes, but also from the information

systems area when considering how to adapt based on

location and behaviour in physical spaces (Not et al.,

1998; Oppermann and Specht, 1999).

3 DIMENSIONS OF ADAPTATION

DTB playback is possible over a broad range of plat-

forms, devices and environments. Also, users with

different physical and cognitive characteristics, pref-

erences and knowledge are expected to use DTBs. For

this reason an automatic and ﬂexible DTB building

platform is a necessity (Carric¸o et al., 2004). Nev-

ertheless, there are situations where it is not possi-

ble to previously identify all the variables governing

the book’s production. Even when the users and sit-

uation share common characteristics, it is impossible

to please everyone. DTBs produced by the DiTaBBu

platform (Carric¸o et al., 2004) were evaluated (Duarte

et al., 2003b) and some of the subjects’ observations

are evidence for having a personalized version of the

books. Examples are given bellow:

• I would like to have the full text of the annotation

shown by default instead of just the subject.

• I would like a sound to signal the start of a new

chapter.

• I would like the start of a new chapter to be dis-

played in the text.

• The annotation frame could be replaced by a link,

that when followed would show an annotations

window. The space currently occupied by the an-

notation frame would be used to display more text.

• The annotation frame could be used to show images

related to the text.

• I would like to have links to web pages related to

the books context.

• I would like for the text to have signals showing

where it has been annotated, instead of having just

an audio signal.

Each of these observations was made by a different

subject, with the exception of the request for a sound

signalling the start of a chapter. This is an indica-

tion that each person would like a personalized book

player.

Considering the blind and print-disabled popula-

tion, which is the primary target audience for DTBs,

we can expect users with very different characteristics

from the test participants. These, in addition to not

being visually impaired, were familiar with computer

interfaces in general. We expect the identiﬁcation of

the aspects to adapt, and the adaptation process, to

be a considerably harder task for the target audience,

thus being an extra motivation for the development of

an adaptive interface for the DTB.

Besides the mentioned “personal preferences”,

other aspects should be contemplated when consid-

ering the creation of an adaptive book. General char-

acteristics include the capabilities of the playback ma-

chine (which input/output devices are available), the

playback environment (for instance, a noisy back-

ground), the users current activities and others that

can be found in common adaptive applications. Tak-

ing into account all these and other variables speciﬁc

to DTBs we identiﬁed some of the elements of the

presentation and interaction that can be adapted, as

well as some of the variables responsible for initiat-

ing the adaptation procedure. The adaptation initiat-

ing variables can be divided into two groups: user re-

lated and environment related. Examples of variables

are:

User related variables - characteristics, knowledge,

preferences, interaction history and current activity.

Environment related variables - interaction de-

vices available, access to media repository,

background noise.

The adaptable components of the book may be sep-

arated into three dimensions: interaction, content and

presentation. Examples of some of the components of

each dimension are:

Interaction - The input/output modalities available

can be enabled or disabled, and used in cooperation

or individually.

Content - Enhancement of the book with the intro-

duction of sound, images, and other available me-

dia, translation of the text, hiding or revealing of

parts of the book.

Presentation - Size and colour of the font used, rear-

ranging of the elements of the book on screen, type

of audio signals used, synchronization units.

Taking into greater account the speciﬁc aspects of

DTB generation and playback, and of its target au-

dience, we can identify the most important book and

user related variables and adaptable components: the

visual impairment level of the user (variable) and syn-

chronization units (adaptable component).

ICEIS 2004 - HUMAN-COMPUTER INTERACTION

148

We can also examine DTB related aspects of the ar-

ticulation between some of these variables and com-

ponents.

• The visual-impairment level has a clear impact on

the presentation of the book. From the size of the

font used, to not using visual presentation at all for

blind users.

• The user preferences, capacities, and past interac-

tions inﬂuence the presentation of the book, by set-

ting the level of enhancement used.

• The interaction history can impact the synchroniza-

tion units. If there is a history of jumping to the

same (or near) word in the book, the synchroniza-

tion units should become gradually ﬁner. In this

way, the ﬁrst jumps would lead to the word’s para-

graph, then to the word’s sentence, and ﬁnally to

the word itself.

• The output devices used can impact the synchro-

nization units. Absence of a visual display leads to

the use of less detailed synchronization units. Af-

ter searching a word in a sound only output envi-

ronment, a jump to the searched word would make

it more difﬁcult for the reader to understand the

“new” context, when compared with a multimodal

environment, in which the narration jump can be

accompanied with the visual presentation of the

surrounding text.

The book characteristics will also play a part in the

adaptation process. For example, when processing a

novel, it wouldn’t make much sense to hide parts of

the text, but when processing an educational book it

could make sense to hide some of the content already

know by the reader.

In order to produce adaptive versions of the DTBs

we propose to evolve the DTB production framework.

The adaptation dimensions are related to adaptations

modules that are responsible for introducing the adap-

tation capabilities in the generated DTB. The modules

and their integration are discussed in the next section.

4 USER INTERFACE

ARCHITECTURE

The design of the presentation and interaction aspects

of a DTB must consider several issues, including the

target user, the execution platform and the reproduc-

tion environment. These and other issues impose con-

straints that the user interface must conform to. If the

reader is print-disabled, both presentation and inter-

action should be done using audio. If the book is to

be read in a PDA, then the presentation and the in-

teractors must adapt to the reduced display size. It

the book is presented in a noisy environment, then the

audio volume must be adjusted, or even discard the

audio output in favour of a visual presentation.

When confronted with the diversity of user char-

acteristics, available devices, and all the other is-

sues controlling the presentation and interaction with

DTBs we can see the impossibility of having an “one-

size ﬁts all” interface. Also, considering the number

of possible combinations of all the elements, it would

be unreasonable to develop a user interface for each

of the possibilities. However, if we are able to estab-

lish a set of relations between the interface generation

controlling aspects and the usable interactors, then we

can create a unique interface whenever the need for it

arises. This interface will be tailored to the character-

istics of the user and the environment, and will be able

to efﬁciently use the available interaction devices.

A framework with these characteristics has several

advantages: ﬂexible creation of user interfaces, allow-

ing an easier interface development process by speed-

ing up the creation-evaluation cycle; development of

reusable presentation and interaction templates in-

dexed by the characteristics of the platform, the target

users and the environment; implementation of multi-

modal behaviour by integration of templates designed

for separate modalities; and adaptation of the inter-

face to changes in the platform, environment or user.

We adopt a model-based approach (Patern

o, 2000)

to the development of our DTB’s user interface. The

UI generation architecture is represented in ﬁgure 1.

Book

Content

Synchronization

Information

Navigation

Information

Platform

Model

Devices

Model

Presentation

Model

User

Model

Environment

Model

Presentable and

Interactable

Book

Audio

Narration

Figure 1: The UI generation architecture.

The interface generation process receives as in-

put the normalized version of the book’s content, the

book’s narration in audio format, the synchronization

between text and audio, and the book’s navigation in-

formation. For details on the creation of these inputs

we refer the reader to (Carric¸o et al., 2004; Duarte

et al., 2003a).

Depending on the resources available in the execu-

tion platform, the interface can be customized by the

user, or even adjust automatically to changes in the

environment or in the platform itself (e.g. the con-

nection or disconnection of new interaction devices

during the book playback). For this to be possible,

A FLEXIBLE INTERFACE ARCHITECTURE FOR DIGITAL TALKING BOOKS

149

the generation environment needs to be “migrated”

to the running environment. A static execution en-

vironment is able to present only previously gener-

ated books, made for speciﬁc combinations of plat-

form and users. In a dynamic execution environment,

changes in the platform, user and environment would

trigger the adaptation of the interface to its new con-

ditions. When discussing the models used in our ar-

chitecture we refer to these two situations as design-

time generation and run-time generation. Generation

in design-time creates a book tailored for speciﬁc ex-

ecution platforms and user groups. Generation in run-

time allows the book to dynamically adapt to changes.

Five models are used to describe the different as-

pects conditioning the development of the interface:

Platform Model - The platform model describes the

various systems capable of presenting a DTB. Ex-

amples are a desktop PC or a PDA. The interaction

devices that are guaranteed to always accompany

the platform are also described in this model. Rel-

evant data that should be described in this model

includes screen size and resolution, and access to

other media sources (for the possibility of obtain-

ing different materials for the book’s presentation,

like images or videos).

Devices Model - The devices model describes all the

devices that can be added to the execution platform,

bringing new modalities into play during the book’s

presentation. Both hardware and software devices

should be included. Examples of hardware de-

vices are a microphone, a braille reader or a braille

keyboard. Possible software devices include text

to speech converters and voice recognizers. This

model can be used in design-time to complement

the platform model and create user interfaces for a

“ﬁxed platform”. However, an increased value can

be obtained by using it dynamically in run-time, re-

acting to changes in the available devices in a way

that allows users to exploit the improved interaction

possibilities.

User Model - The user model describes relevant user

preferences and characteristics. Users can be

grouped by their characteristics and abilities, and

a design-time user interface generation system cre-

ates different interfaces for speciﬁc groups. Tak-

ing into account the particular domain of applica-

tion, relevant characteristics for identifying these

user groups include the visual ability of the user,

the age, etc. A dynamic version of the interface

can adapt to perceived changes in the user’s char-

acteristics and preferences. Examples include the

user’s favourite spatial layout and disposition of the

on-screen components, or the user’s preference for

visual or audio interaction.

Environment Model - The environment model de-

scribes the environmental characteristics that can

have an impact on the book’s presentation and in-

teraction aspects. An example is the ambient noise

of the reproduction environment.

Presentation Model - The presentation model de-

scribes all the components available for present-

ing and interacting with the book. Examples of

components to include in a presentation are the

table of contents, the annotations, images, videos

or audio clips to enrich the reading experience,

the book content itself, and others. Each compo-

nent has a set of implementation templates avail-

able. For each component there exists a template

for each platform where it is possible to execute

the book’s interface, a template for each device that

can be added to the execution platform, a template

for each identiﬁed user group, and a template for

each execution environment. From the informa-

tion in each of the models, a set of rules selects

the proper templates, and combines them to form a

presentable and interactable book. The content to

ﬁll the templates is taken from the inputs to the sys-

tem, i.e. the book content, the audio narration, and

the navigation and synchronization information.

Several of the models can be used to generate the

book in run-time, thus offering the possibility of adap-

tation to changing conditions. To generate a book in

design-time, the user interface designer simply has to

select the platform, devices, user, environment and

presentation models that are most appropriate to the

forecast usage situation. This creates a book that is

“optimal” (according to the designer’s criteria) for the

chosen situation, but that has no adaptation possibili-

ties.

To explore an adaptable interface, the execution

platform must be augmented with generation possi-

bilities. For example, if a PC is used as the execu-

tion platform, with the book being displayed in a web

browser, one way to introduce adaptation capabilities

in the interface is to install a web server with the ca-

pacity to run the generation scripts.

The level of adaptation of the interface can be se-

lected during the design of the interface. By level of

adaptation we mean the range of events to which the

interface should adapt to. This is achieved by includ-

ing one or more models in the executable presenta-

tion. The only model that is not available for includ-

ing in the adaptable interface is the platform model.

This is justiﬁed by the profound changes needed in

the interface when migrating to different platforms,

and the eventuality of being impossible to execute an

adaptable interface in the new platform. All the other

models may be included in the execution platform. If

the designer is unsure of the availability of some de-

vices in the execution platform, the inclusion of the

devices model allows for the adaptation of the inter-

face whenever a device is connected or disconnected

ICEIS 2004 - HUMAN-COMPUTER INTERACTION

150

from the platform. If the designer wishes the interface

to adapt to the behaviour of the user (for instance, if a

user is always requesting to see images enriching the

presentation, the interface should start displaying im-

ages by default to that user), then the inclusion of a

user model may introduce that feature.

An interface without any of the models available

in run-time would be a ﬁxed interface. An interface

with all the models available would be a fully adapt-

able interface. The designer should choose the mod-

els which will be transferred to the run-time interface,

based on her knowledge of the target users, platforms

and environment.

5 CONCLUSIONS

In this paper we have presented the motivations for

the development of an adaptable interface for DTBs.

Variables governing the adaptation and the possible

adaptable components have been identiﬁed, and sev-

eral relations between variables and components pre-

sented.

A model-based architecture for the creation of ﬂex-

ible multimodal user interfaces for DTBs was pro-

posed. This architecture is the base for further de-

velopments, which target the creation of an adaptable

user interface. The introduction of some of the inter-

face development models in the run-time version of

the interface will allow for the possibility of adaptive

behaviour, if supported by the interface presentation

platform.

REFERENCES

ANSI/NISO (2002). Speciﬁcations for the digital talking

book. http://www.niso.org/standards/resources/Z39-

86-2002.html.

Brusilowsky, P. (2001). Adaptive hypermedia. User Mod-

eling and User-Adapted Interaction, 11(1-2):87–110.

Carr, B. and Goldstein, I. P. (1977). Overlays: A theory of

modelling for computer aided instruction. Ai memo,

MIT, Cambridge, MA.

Carric¸o, L., Duarte, C., Lopes, R., Rodrigues, M., and

Guimar

aes, N. (2004). Computer-Aided Design of

User Interfaces IV, chapter Building Rich User Inter-

faces for Digital Talking Books. Kluwer Academic

Publishers. Accepted for publication.

Carric¸o, L., Guimar

aes, N., Duarte, C., Chambel, T., and

Sim

oes, H. (2003). Spoken books: Multimodal inter-

action and information repurposing. In Proceedings

of HCII’2003, International Conference on Human-

Computer Interaction, pages 680–684, Crete, Greece.

Christian, K., Kules, B., Shneiderman, B., and Youssef,

A. (2000). A comparison of voice controlled and

mouse controlled web browsing. In Proceedings of

ASSETS’00, pages 72–79, Arlington, VA.

Duarte, C. and Carric¸o, L. (2004). Identifying adaptation

dimensions in digital talking books. In Proceedings of

IUI’04, Madeira, Portugal.

Duarte, C., Carric¸o, L., Chambel, T., and Guimar

aes, N.

(2003a). Producing DTB from audio tapes. In

Proceedings of ICEIS’03, volume 3, pages 582–585,

Angers, France.

Duarte, C., Chambel, T., Carric¸o, L., Guimar

aes, N.,

and Sim

oes, H. (2003b). A multimodal inter-

face for digital talking books. In Proceedings of

WWW/INTERNET 2003, Algarve, Portugal. Accepted

for publication.

Gazzaniga, M. S., Ivry, R. B., and Mangun, G. R. (1998).

Cognitive Neuroscience - the Biology of the Mind. W.

W. Norton & Company.

Kobsa, A. (2001). Generic user modeling systems. User

Modeling and User-Adapted Interaction, 11(1-2):49–

63.

Kok, A. (1991). A review and synthesis of user modeling

in intelligent systems. The Knowledge Engineering

Review, 6:21–47.

Langley, P. (1999). User modeling in adaptive interfaces.

In Proceedings of the 7th International Conference on

User Modeling, pages 357–370, Banff, Alberta.

NISO (1999). Document navigation features list.

http://www.loc.gov/nls/z3986/background/navi-

gation.htm.

Not, E., Petrelli, D., Sarini, M., Stock, O., Strapparava, C.,

and Zancarano, M. (1998). Hypernavigation in the

physical space: Adapting presentation to the user and

to the situational context. New Review of Multimedia

and Hypermedia, 4:33–45.

Oppermann, R. and Specht, M. (1999). Adaptive informa-

tion for nomadic activities: A process oriented ap-

proach. In Proceedings of Software Engineering’99,

pages 255–264, Walldorf, Germany.

Oviatt, S., Darrell, T., and Flickner, M. (2004). Multimodal

interfaces that ﬂex, adapt, and persist - introduction.

Commun. ACM, 47(1):30–33.

Oviatt, S. L., Cohen, P. R., Wu, L., Vergo, J., Duncan, L.,

Suhm, B., Bers, J., Holzman, T., Winograd, T., Lan-

day, J., Larson, J., and Ferro, D. (2000). Designing

the user interface for multimodal speech and gesture

applications: State-of-the-art systems and research di-

rections. Human Computer Interaction, 15(4):263–

322.

Patern

o, F. (2000). Model-Based Design and Evaluation of

Interactive Applications. Springer Verlag.

Rich, E. (1989). Stereotypes and user modelling. In Kobsa,

A. and Wahlster, W., editors, User Models in Dialog

Systems, pages 35–51. Springer Verlag, Berlin.

Van Buskirk, R. and LaLomia, M. (1995). A comparison of

speech and mouse/keyboard gui navigation. In Pro-

ceedings of CHI’95, Denver, CO.

A FLEXIBLE INTERFACE ARCHITECTURE FOR DIGITAL TALKING BOOKS

151