This extended abstract is a contribution to the User Modeling for Accessibility Online Symposium of 15 July 2013. The contents of this paper was not developed by the W3C Web Accessibility Initiative (WAI) and does not necessarily represent the consensus view of its membership.

Increasing the Flexibility of Accessibility Modelling Through the Use of Semantic Relationships

1. Problem Description

The accessibility solution that is appropriate for an individual in a given situation may or may not be a dedicated assistive technology. Adaptations can work as micro-ATs, providing personalisation that increases the range of users who are able to access content (Vanderheiden, 2008). As they are intended for customisation, many adaptations are not labelled as accessibility options. As they form a crucial role in ensuring accessibility, adaptations and device-specific settings are increasingly being included in profiles describing accessibility needs.

The variety of interaction paradigms results in the need for specificity in technology profiles to capture the nuances of each particular device, interface or control. As users are frequently interacting with multiple technologies there is also a need for a user profile that is generic enough to be transportable, whilst specific enough to respond to technology nuances. One solution to this need is to use the Semantic Web to model adaptations (and other accessibility solutions) in terms of human capabilities (Atkinson et al. 2012). This paper focuses on providing a vocabulary that moves from device-specific to device-agnostic profiling through the logical structuring of profiles. The approach improves flexibility, during both profile acquisition/maintenance and the matching process. Dynamic comparison of profiles at varying (appropriate) levels of granularity, allows discovery and remedy of accessibility issues to be performed efficiently.

2. Background

Accessibility problems are discovered via the comparison of profiles describing the user and the technology they wish to use.

The Common Access Profile (CAP) provides the grounding for ISO 24756 and allows direct comparison between the needs and capabilities of users, systems and their environment (Fourney, 2006). It views accessibility in terms of channels of communication that are facilitated by capabilities and constricted by impairments. CAPs are constructed from a series of Interacting Components which are able to either input or output via specified modalities. However the standard demands a fixed vocabulary of modalities resulting in rigidity.

ISO 24751 deals with the provision of individualised adaptability and accessibility in the context of e-learning. It is a three part standard which describes the user and application profiles separately (ISO/IEC 24751-1:2008). Currently under revision, there is discussion as to its revised vocabulary and content. Profiles could be described either in terms of device-focused preferences (the current language used) or human-focused capabilities. Preferences refer to device or software settings that can be specified to improve accessibility (e.g. font size). As they are device-specific, transportability of a user’s profile is dependent on mappings between all related preferences across all devices. Capabilities refer to the human abilities used in interaction (e.g. visual acuity). Unlike preferences they provide a single, static vocabulary against which any device setting may be mapped (Atkinson et al. 2010). The full gamut of human capabilities is available for use from low level ergonomic (Peissner, 2012) to higher level functional capabilities (Billi, 2006).

The Semantic Web is a collection of standards that allow data to be “shared and reused across application, enterprise and community boundaries.” (W3C, 2013) By storing data in terms of its relationships to other data, networks can be formed to produce machine readable information. The Semantic Web has successfully been applied to user modelling in several projects including the Composite Capabilities/Preference Profiles (CC/PP) (Klyne, 2004).

3. Approach

The approach taken in this paper builds on the established profiling techniques above by using a series of semantic relationships within a framework which logically describes both users and technology using symmetrical hierarchical structures. Needs and capabilities are described at various granularities, with human capabilities used to allow direct matching between profiles at the appropriate abstract level. This approach reduces the rigidity suffered through the simple application of either of the standards mentioned above.

Profiles are built up using a nested series of actors. An actor may possess several sub-actors describing its components (‘HAS’). Each actor represents a user, piece of technology, or environment (or part thereof) with capabilities to interact, be interacted with, or interfere with interaction (‘CAN’). Capabilities are described in terms of the abilities or needs they expose, in order to allow channels of communication to be built and investigated between actors. Capabilities share the same hierarchical structure as actors with higher-level abilities being inferred by the presence of lower-level abilities (‘INFERS’) or being reliant on the presence of other abilities (‘REQUIRES’).

As an example, a user and their mobile phone would both be represented as actors with capabilities representing their high level needs and abilities. If these cannot be directly compared, their sub-actors can be queried to expose lower-level capabilities until a direct comparison can be made. The actor representing the user may have sub-actors representing their hands, eyes, ears etc. The mobile phone may have sub-actors representing the case, microphone, speaker, buttons, touch-screen, and the applications it contains (an application in turn being comprised of interface elements).

Each sub-actor is an actor in its own right. As a sub-actor of the phone, the application’s capabilities would be attributed to the phone (and as sub-actors of the application the same would be true of the interface elements).

4. Challenges

Semantic web technologies are not yet fully standardised, meaning there are still difficulties to be overcome before they are able to be widely adopted. Research on accessibility profile development is also proving successful at a theoretical level, however outstanding challenges include:

Data Acquisition – The research described is focused on the description, storage and use of profile data. The acquisition of the information required to populate a profile is two-fold: (1) initial population may require a bootstrapping procedure, and (2) information would need to be kept up-to-date. For example, ISO 24756 relies on the user (or other human agent) to create and update CAPs. However for mass adoption, the user cannot be relied upon either because (1) they may be unwilling, or (2) the information they provide may be unreliable. The further research section addresses the potential for solutions to this challenge.

Context – Any data collected is constrained by the context within which it is collected (Peissner, 2012). This concern can be addressed via inference back to human capabilities and the ability to model the environment as an actor. Any interference (e.g. background light or noise) will be stored as a capability in a format compatible with the user’s abilities to perceive them. The ability to hear at a certain volume given a quantified amount of background noise allows the inference of the user’s ability in other contexts.

5. Outcomes

The research presented in this paper provides a flexible method of profiling users and technology, enabling direct matching between abilities and needs. The relationships allow capabilities defined in various open and proprietary standards to be used together. Where a match cannot be found in terms of device preferences, the graph of capabilities can be searched recursively to find a point where comparable abilities are available. This is achieved in part through the use of a nested series of actors, allowing the granularity of assessments to be targeted.

Assistive technology matching can be performed via the speculative augmentation of actors with representative sub-actors. Software can be modelled on different devices by placing its highest representative actor within the actor describing each device. A hearing-aid or adaptation providing text-to-speech functionality would both be described in terms of the way they re-routed and transformed the existing channel of communication.

6. Future Research

The application of the research contained within this paper requires the availability of profile data for both users and technology. Given the challenge of data acquisition described above, the framework is reliant on the availability of agents which are able to populate user profiles. The emphasis for the provision of technology profiles would be placed on developers and manufacturers, or could be championed by early adopters.

With appropriate data, the framework would be able to provide predictions for adaptations and changes to existing setting that would be beneficial to the user. Further research is needed to understand the reactions and preferences of users with regards to a system that provided this functionality. A delivery system would be required, both to facilitate the personalisation of existing settings and the procurement of additional adaptations. The level of autonomy with which a system would be allowed to function would clearly depend on the individual.

References