This item was submitted to Loughborough’s Institutional Repository
(https://dspace.lboro.ac.uk/) by the author and is made available under the
following Creative Commons Licence conditions.
For the full text of this licence, please go to:
http://creativecommons.org/licenses/by-nc-nd/2.5/
Towards Ubiquitous Accessibility: Capability-based
Profiles and Adaptations, Delivered via the Semantic Web
Matthew Tylee Atkinson Matthew J. Bell Colin H. C. Machin
Department of Computer Department of Computer Department of Computer
Science Science Science
Loughborough University Loughborough University Loughborough University
Leicestershire Leicestershire Leicestershire
LE11 3TU, UK LE11 3TU, UK LE11 3TU, UK
M.T.Atkinson@lboro.ac.uk M.J.Bell@lboro.ac.uk C.H.C.Machin@lboro.ac.uk
ABSTRACT
The continuing proliferation of mobile devices, content and
applications presents barriers to the mainstreaming of As-
sistive Technologies (ATs), despite their potential utility for
users in demanding situations or with minor-to-moderate
impairments. We have previously proposed that user profil-
ing based on human rather than machine-oriented capabil-
ities, coupled with a shift from conspicuous ATs to consid-
ering a broader range of adaptations presents opportunities
for platform and AT vendors to support many more users.
However there has not been a standard, consistent and, most
importantly, straightforward way to deliver these benefits.
We propose that this delivery gap can be bridged by using
the semantic web and related technologies, so the potential
benefits of the capability-based approach may be realised.
Categories and Subject Descriptors
H.5.4 [Information Systems]: Hypertext/Hypermedia—
User issues; K.4.2 [Computing Milieux]: Computers and
Society—Assistive technologies for persons with disabilities
Keywords
Adaptation, Capability, OWL, Profile, RDF, Semantic, User
1. INTRODUCTION
Users have an increasingly large choice of ways to access
digital content and are spreading their time across multi-
ple devices. We (and others, as discussed in [5]) proposed
that an adaptive approach to interface design allows the user
more control to personalise their interaction providing bet-
ter ubiquitous [6] accessibility.
Currently the assistive technology landscape is vast and
users may find it difficult to find the right assistive tech-
nology for them. When a solution is found it then has to
be checked not only for compatibility with the system, but
that it does not disrupt the user through an unintended
side-effect. For the technical w
ork-package of the wider, so-
ciotechnical Sus-IT project
1
we have employed an adaptive
support system allowing assistance to be offered that is ap-
propriate to the user given their capabilities [1].
In order to facilitate this kind of system, reasoning must
be carried out as close to the problem domain as possible
and flexible storage needs to be implemented that is able to
aggregate the often disparate data needed to make decisions
and match assistive technology with a user in need. Mod-
elling the user based on their capabilities, a holistic storage
framework was developed [2].
1.1 Contributions
Here we argue that capability-based reasoning about adap-
tations, delivered via the semantic web, will both open up
accessibility to the wider population and enable many stake-
holder organisations to benefit.
Due to its rise in popularity and potential to provide cross-
platform services, the semantic web has been chosen as the
delivery method for the proposed storage system. In partic-
ular the Resource Description Framework (RDF) combined
with the Web Ontology Language (OWL) have been chosen
due to their flexible and descriptive nature, as well as the
potential to use inference capabilities.
1.2 Trends and Opportunities
Three converging trends create the opportunity for capability-
based reasoning, delivered via the semantic web, to bring
improvements in accessibility for the wider population.
• Ubiquitous and differentiated devices. (Requires sup-
porting a more diverse user group and profiles that are
more portable across devices.)
• Service-based content and application delivery. (Re-
quires accessibility delivery to be web-compatible.
2
)
• More adaptable/mutable applications, e.g. mash-ups.
(Requires smaller, “Micro-ATs” [6] which, along with
other customisation options and traditional ATs, we
call adaptations.)
A wide range of research projects work in these areas;
Table 1 provides an overview of the goals of some of these.
1
http://sus-it.lboro.ac.uk/
2
A number, such as ATbar (http://atbar.org/) already
are, but are not easily discoverable by novice users.
Permission to make digital or hard copies of all or part of this work for
personal or classroom use is granted without fee provided that copies are
not made or distributed for profit or commercial advantage and that copies
bear this notice and the full citation on the first page. To copy otherwise, to
republish, to post on servers or to redistribute to lists, requires prior specific
permission and/or a fee.
W4A2012 - Communication, April 16-17, 2012, Lyon, France. Co-Located
with the 21st International World Wide Web Conference.
Copyright 2012 ACM 978-1-4503-1019-2 ...
$5.00.
© ACM, 2012. This is the author's version of the work. It is posted here by permission of ACM for your personal use.
Not for redistribution. The definitive version was published in Proceedings of the International Cross-Disciplinary
Conference on Web Accessibility (W4A), Lyon, France (April 16- 17, 2012): http://dx.doi.org/10.1145/2207016.2207020
Table 1: Overview of the goals of some related projects
Includes Based on human Easy user self- Roaming (Subtle) Adaptation
Project hardware capabilities identification profiles adaptations discovery
SNAPI
1
CC/PP
2
GUIDE
3
GPII
4
Yes
No
Yes
No
No
No
Yes (not standardised)
Possible
Yes (smartcard)
Out of scope
Possible
Possible
Possible
Possible
Possible
Yes
Hardcoded
Hardcoded
Yes (inc. dev. toolkit)
Yes (Micro-ATs)
Out of scope
Out of scope
No
Yes
Sus-IT No Yes Possible Yes Yes Yes
Note: as the projects’ goals differ, this is an overview of the field; not a like-for-like comparison: 1: http://snapi.org.uk/
2: http://www.w3.org/TR/2004/REC-CCPP-struct-vocab-20040115/ 3: http://guide-project.eu/ 4: http://gpii.net/
2. PROPOSALS
2.1 Human Capabilities and Reasoning
Device functionality changes rapidly, e.g. mice had scrolling
wheels added to them, but are now being supplanted in
some areas by touch-based input. In order to maintain a
user profile that can be of use across a range of devices and
applications, it is necessary to store that profile in terms
of human capabilities, using ratified standards such as ICF
from the World Health Organisation [7] and associated stan-
dard medical units of measure, rather than machine-specific
quantities (such as pixel depth). Recording details about the
user’s fine motor dexterity can tell us how likely it is they
will be able to use a mouse wheel or form certain touch
gestures (as discussed in [1]).
A person’s visual acuity enables us to determine a mini-
mum sensible font size for any device, at any resolution or
screen size, given an assumed or pre-measured reading dis-
tance. The user’s visual acuity need not be actively probed
by the system; it can be estimated based on the known phys-
ical size of text on a given device. The goal is to progres-
sively refine estimates for the boundaries of a user’s capabil-
ity range in order to suggest appropriate assistance–not to
make abrupt and noticeable adaptations–so a passive, “ball-
park” approach is preferable to one that is more accurate
but constantly questions or tests the user. (Bootstrapping
is discussed in [5, sec. 4.3].)
Cognitive problems are, inherently, more challenging to
resolve as they require the co-operation of content authors
to a much higher degree. These are touched upon later.
2.2 Adaptations: Generalised ATs
Adaptations, as proposed in [5, sec. 4] and [1, sec. 3], are
generalised ATs. They range from sometimes in-built cus-
tomisations such as the base font size for GUI widgets (or
web content), through Vanderheiden’s “Micro-ATs” [6], to
monolithic ATs such as alternative keyboards or screen read-
ers. By considering the whole range of possible adaptations,
suggestions for assistance can be made for people experienc-
ing transient impairments. A cultural, as well as technical,
expectation of adaptivity in systems can help ensure their
architecture is sufficiently open for more specialist ATs [6].
The mechanism of effecting changes may be platform-
specific (e.g. using existing accessibility APIs) or agnostic
(e.g. using the DOM, for content). The DOM is a prime
example of a mature, simple but powerful interface enabling
many types of adaptations to a range of content.
3
However,
3
As used by (X)HTML and GUI toolkits such as Mozilla’s
XUL (http://developer.mozilla.org/en/XUL).
the reasoning process for matching capabilities to adapta-
tions needs to know only the following in order to make
suggestions.
• Capability: requirements of the situation and content;
levels of the user; capacity of the adaptations.
• Side-effects of the adaptations; essentially the degree
to which the adaptation obstructs the flow of informa-
tion from the device (e.g. zooming the screen reduces
the amount that can be seen, resulting in a capability
burden on motor and cognition, due to panning).
2.3 Semantic Web Delivery
RDF is a set of web standards for expressing data and
metadata by making statements in the form of triples. OWL
provides language to impose rules on the data, extending
RDF to allow new statements to be inferred (via rules) from
existing ones. RDF has been successfully applied to user
modelling [4] and forms the basis of CC/PP.
A strength of this approach is its ability to draw upon
metadata. Compiling a holistic profile for a user using a tra-
ditional “static” method would be difficult, as even if a com-
plete profile could be gathered in a timely manner, it would
quickly become out-of-date as the environment changes or
the user gains experience. More appropriately a user profile
will be gradually assembled from data provided both by the
user and the system they are using; as new and changing
capabilities are captured, the profile is kept up-to-date. Is-
sues raised when relying on a static profile become powerful
opportunities when viewed through a semantic lens.
With semantic profiles, available data are processed along
with associated metadata, with the reasoning process con-
tinually assessing their worth to provide a confidence judge-
ment for any decisions made. When data are deemed “out-
of-date” (either due to the passage of time, or conflicting
capability readings arising) they can be archived and pat-
terns can be inferred to plot capability change, e.g. decline,
allowing pre-emptive accessibility solutions to be invoked.
This can prevent the user from suffering the loss of service
(and associated threat of abandonment) typical when reac-
tive support is provided. If data are missing altogether, it
may be possible to use inference from generic profiles aggre-
gated across all users, or those with similar capabilities.
Existing standards can be used and extended as necessary;
the tree structure of WHO’s ICF [7] (which runs from gross
modalities of interaction to specific individual capabilities)
makes it ideal, as content, adaptations and even users can
be marked up to as fine a level of granularity as possible, or
as justifiable by manufacturers and developers, whom may
adopt the approach incrementally.
© ACM, 2012. This is the author's version of the work. It is posted here by permission of ACM for your personal use.
Not for redistribution. The definitive version was published in Proceedings of the International Cross-Disciplinary
Conference on Web Accessibility (W4A), Lyon, France (April 16- 17, 2012): http://dx.doi.org/10.1145/2207016.2207020
Figure 1: Architecture overview.
2.4 Burden on Developers and Authors
A common view amongst development or authoring or-
ganisations is that implementing accessibility will not pro-
duce sufficient return on investment. As researchers, we are
aware that improving accessibility can improve the expe-
rience for all users, particularly in mobile (potentially de-
manding or adverse) situations. Unfortunately, even those
platform vendors that have ATs built into their products
(e.g. Apple’s iPhone
4
) segregate these into “Accessibility”
settings, of which most users remain unaware. Further,
when these ATs are employed, they often drastically change
the way the device must be used (zooming in and panning
the screen rather than just increasing the size of widget text
and reflowing the interface, for example).
The ubiquitous accessibility infrastructure proposed here
and depicted in Figure 1 is designed to support discovery
of appropriate adaptations for most people out-of-the-box.
Adopting it would also enable specialist ATs for those with
recognised disabilities, should they be available for the plat-
form, to function more seamlessly (as hooks into the sys-
tem would be provided to them). Further, as it is based
on existing ratified standards, it would not require as much
effort to implement as the (complimentary) existing accessi-
bility APIs that are already present on most contemporary
devices—and it would improve the return on the investment
in those APIs by enabling more people to discover adapta-
tions that use them.
Platform vendors must include the capability reasoning
library [5, sec. 5.1] in their operating system. They
must also mark up the capability requirements of the
in-built features of their system, and capabilities ad-
dressed by any ATs it includes. This can be done to
any level of granularity as the ICF is tree-structured.
Devices must have capability requirements mark-up.
Application and Content authors may optionally mark
up their content, to any level of granularity. By default
the reasoning process will fall back to reasoning about
the suitability of the modalities of the elements in the
document. Alternative content may be required to
support cognitive impairments as presentation adap-
tations may be insufficient.
4
http://www.apple.com/accessibility/iphone/
Adaptations must enumerate the capabilities they assist.
Due to the trends discussed in section 1.2, it is surmised
that this extra effort will be seen to be cost-effective in the
near future. Our reasoning processes provide drop-in candi-
dates for the proposed GPII’s matchmaker and preferences
storage components
5
, thus increasing their potential useful-
ness in the eyes of platform and AT vendors.
3. APPLICATIONS AND ADVANTAGES
The theme of the proposed technique is to build bridges
between existing recognised standards and forms of assis-
tance, in order to make them more discoverable to users.
3.1 Using Existing Standards
The key bridge to be built (described in [1]) is between
the established standards for human capabilities, such as
WHO’s classification of functioning, disability and health [7]
and those for the technical adjustments that can be made to
machines, such as ISO 24751 [3]. Existing profiling methods
(e.g. SNAPI; CC/PP; [4]) tend to express profiles at the
more technical end of the spectrum, which is not as portable
to new types of device as capability data. The following
sections are all based on this central reasoning technique.
3.2 User–Adaptations
By considering users’ preferences as well as the potential
gap between the capabilities required by the content, device
and situation and those possessed by the user, we can sug-
gest appropriate adaptations (settings changes; micro-ATs
or ATs). “Preferences” are the trade-offs the user is will-
ing to make when an adaptation is made, as adaptations
often impede the flow of information in order to make that
information that is presented clearer in some way.
When presented with a choice, would the user prefer, e.g.,
2D scrolling, keeping the normal visual layout, or would they
prefer 1D scrolling, at the cost of re-flowing the layout (mak-
ing it different to that experienced by other users). These
trade-offs are encoded in the abstract, allowing the reasoner,
in this example, to determine that 2D scrolling causes higher
cognitive and motor load. If the user had dexterity prob-
lems too, then the re-flowed 1D scrolling alternative would
be more highly recommended than 2D scrolling.
3.3 User–Device
We can generalise this to a situation in which the user
is free to choose the most suitable device. For example, a
new smartphone could be recommended based on the users’
and devices’ relative capability match. This could present
an incentive for retailers to support the proposed user profil-
ing technique–and would not be possible unless the profiles
were based on existing ratified and impartial standards such
as [7], nor available on the web in a ubiquitous interchange
format such as RDF.
3.4 Product Vendor–AT Developer
As noted by Vanderheiden [6], an open market for adap-
tations could be created (an effective “Accessibility App
Store”—though we submit that it should be tightly inte-
grated into the host platform, rather than segregated, for
the reasons of user awareness given in section 2.4). AT de-
velopers could increase their potential market massively by
5
http://gpii.net/components
© ACM, 2012. This is the author's version of the work. It is posted here by permission of ACM for your personal use.
Not for redistribution. The definitive version was published in Proceedings of the International Cross-Disciplinary
Conference on Web Accessibility (W4A), Lyon, France (April 16- 17, 2012): http://dx.doi.org/10.1145/2207016.2207020
moving to micro-ATs based on user capabilities. Platform
vendors would intrinsically gain improved access for a more
diverse user population, which would improve as more adap-
tations are developed. Ethical matters around recommen-
dations and profile security are discussed in [5, sec. 4.4].
4. CONCLUSIONS
This paper proposes the union of three techniques to aid
the provision of accessibility solutions. AT provision should
be made via adaptations rather than the traditional conspic-
uous and monolithic ATs. This both increases the impact
the chosen adaptation has, due to smaller ATs being more
focused on specific problems, and decreases undesirable im-
pact through conflicts with other hardware and software.
Storing user profiles and marking up content and adapta-
tions in human capability terms allows problem-centred rea-
soning; finding the right adaptation for the problem becomes
a comparison between needs and abilities, albeit with the
discussed preference (trade-off), and relevant device, con-
straints. The possible solutions may be ranked against these
constraints to evaluate the potential benefit available.
Using the semantic web as the the delivery mechanism
for the established capability standards, as well as profiles
and adaptation catalogues provides a flexible descriptive lan-
guage that is easily transported, platform-independent and
can inherently make use of inference, as described.
Should platforms also adopt integrated adaptation direc-
tories/marketplaces, appropriate adaptations may be sug-
gested for any given device and situation. The likelihood
of technology abandonment may be significantly reduced as
cross-device optimisation of users’ interactions is possible,
based upon a dynamic portable personal profile.
4.1 Illustrative Example
Imagine a user with fine motor dexterity problems. Record-
ing an ability to use a mouse, but not its scroll wheel, is
too device-specific. Given the capability-centred approach
we would store information regarding the capabilities of the
user’s finger. The inability is indicative of a finger dexterity
problem reduced capacity in fine motor skills. On trying to
use a public multi-touch terminal, the user may find their
reduced dexterity a problem, as it may preclude using pinch-
to-zoom gestures. Although the user may never have used a
multi-touch device before, inference can be made from their
lack of fine motor capability that the pinch gesture could
be unattainable. A zoom widget, such as a large slider bar,
can be provided for the user. On a small device such as a
tablet, where screen space is at a premium, this would not be
provided for most users. On a public information terminal
there is likely more space, but given the popular design aes-
thetic of minimising screen clutter, an explicit zoom widget
may not have otherwise been provided. Further: if a user is
not known to have experience with multitouch interaction—
therefore lacking the appropriate mental model—the termi-
nal can be adjusted to offer explanation.
It is the combination of capability based profiling, seman-
tic storage and adaptive accessibility that make this ap-
proach so powerful.
4.2 Future Work
We developed a capability-based adaptation system for
Sus-IT (with acceptance testing complete and longitudinal
testing ongoing). We now wish to: (1) prepare this for de-
livery using the techniques proposed here, with the goal of
wider adoption and (2) improve tools for developing cogni-
tive adaptations.
In order to support cognitive adaptations, such as context-
and-capability-sensitive help for applications, more mutable
applications are needed (to provide hooks for such adapta-
tions), as is additional work from content authors. Examples
of such adaptations include reminders or tutorials on how
to use programs.
6
It would be possible to employ a crowd-
sourcing approach to the creation, storage and capability-
based recommendation of such help material, though more
accessible authoring tools will be required to lower the bar-
riers to creating them.
5. ACKNOWLEDGMENTS
This paper emerges from research conducted as a part of
the Sus-IT Research Project, led by Leela Damodaran and
Wendy Olphert, funded by the UK ESRC’s New Dynamics
of Ageing Programme (Grant Number RES-353-25-0008).
6. REFERENCES
[1] M. Atkinson, Y. Li, C. Machin, and D. Sloan. Towards
accessible interactions with pervasive interfaces, based
on human capabilities. In K. Miesenberger, J. Klaus,
W. Zagler, and A. Karshmer, editors, Computers
Helping People with Special Needs, volume 6179,
chapter 27, pages 162–169. Springer Berlin Heidelberg,
2010.
[2] M. J. Bell and C. H. C. Machin. A framework for
adaptive communication design. In SIGDOC ’09:
Proceedings of the 27th ACM international conference
on Design of communication, pages 45–50. ACM, 2009.
[3] International Organization for Standardization.
Information technology – individualized adaptability
and accessibility in e-learning, education and training.
ISO Standard, 2008.
[4] I. Palmisano, D. Redavid, L. Iannone, G. Semeraro,
M. Degemmis, P. Lops, and O. Licchelli. A rdf-based
framework for user profile creation and management. In
B. Gabrys, R. Howlett, and L. Jain, editors,
Knowledge-Based Intelligent Information and
Enginee ring Systems, volume 4253 of Lecture Notes in
Computer Science, pages 606–613. Springer Berlin /
Heidelberg, 2006.
[5] D. Sloan, M. T. Atkinson, C. Machin, and Y. Li. The
potential of adaptive interfaces as an accessibility aid
for older web users. In W4A ’10: Proceedings of the
International Cross Disciplinary Conference on Web
Accessibility, pages 1–10. ACM, 2010.
[6] G. C. Vanderheiden. Ubiquitous accessibility, common
technology core, and micro assistive technology:
Commentary on
`
‘computers and people with
´
disabilities’. ACM Trans. Access. Comput., 1:10:1–10:7,
October 2008.
[7] World Health Organization. International classification
of functioning, disability and health (ICF).
http://www.who.int/classifications/icf/, 2001.
6
These could be automatically adapted so that they are ren-
dered appropriately for some users, though alternative con-
tent may be required for those with cognitive impairments.
© ACM, 2012. This is the author's version of the work. It is posted here by permission of ACM for your personal use.
Not for redistribution. The definitive version was published in Proceedings of the International Cross-Disciplinary
Conference on Web Accessibility (W4A), Lyon, France (April 16- 17, 2012): http://dx.doi.org/10.1145/2207016.2207020
