Preamble

The [United Nations] Convention on the Rights of Persons with Disabilities and its Optional Protocol were adopted by the United Nations General Assembly on 13 December 2006, and opened for signature on 30 March 2007. On 30 March, 81 Member States and the European Community signed the Convention, the highest number of signatures of any human rights convention on its opening day. 44 Member States signed the Optional Protocol, and 1 Member State ratified the Convention. The Convention was negotiated during eight sessions of an Ad Hoc Committee of the General Assembly from 2002 to 2006, making it the fastest negotiated human rights treaty. The Convention aims to ensure that persons with disabilities enjoy human rights on an equal basis with others. [UN Enable, 2008a]

By March 31, 2008, there were 126 signatories to the United Nations Convention, 71 signatories to the Optional Protocol, 18 ratifications of the Convention and 11 ratifications of the Optional protocol (not including Australia) (UN Enable, 2008a). In an information era, everyone should have, one way or another, an equal right to information if they are to participate equally in the information age. The general aim of the new United Nations convention is to ensure that people with disabilities are treated inclusively as are other groups of people identified in earlier conventions. In particular, this convention calls for inclusive access to information and communications for people with disabilities, and specifies a number of situations in which these rights must be enforced, including for work, entertainment, health, politics and more (UN, 2006).

This thesis has, at its heart, the idea that inclusive treatment of people eliminates the need for special considerations for people with disabilities. This is derived from the social model of disability (Oliver, 1990b). First, it attends to the limits on people's abilities to participate in society rather than on any medically defined 'defect' they may be considered to have. Secondly, it supports equally able-bodied people who for one reason or another cannot participate equally.

The social model of disability spreads responsibility for inclusion across the community. This research aims to enable continuous, distributed, community effort to make the World Wide Web inclusive.

For a decade, efforts to make the Web accessible have focused on following, or otherwise, a set of guidelines that have come to be treated as specifications. These guidelines, based on an accessibility model that assumes a single resource could be made available to everyone (universal design), have proven inadequate to ensure accessibility for all. Recent estimates of the accessibility of the Web are as low as 3% (e-Government Unit, UK Cabinet Office, 2005). The research indicates that the current level of reliance on the universal accessibility model is flawed.

If a user is blind, eyes-busy or using a small screen, instructions about how to get from one place to another presented as a map may be incapable of perception while a text version that can be read out and heard would be perceptible. Providing a text description of travel routes is an example of an accessibility improvement for a map. Managing the map and the new version so that it is associated with the map, and discoverable at the same time as the map, is what catalogue records or metadata can do for digital objects.

The research advocates a process to support ongoing incremental improvement of accessibility. This depends upon efficient management and description of distributed resources and their improvements, and descriptions of them. These can be matched to individual's needs and preferences. The research elaborates what is called AccessForAll metadata (Nevile & Treviranus, 2006), a framework for descriptions of resources and resource components. AccessForAll metadata provides for a common set of descriptions so that they can be shared, so they will interoperate across description protocols, and so they can be used by computers to automatically match resources to users' needs and preferences. AccessForAll metadata includes provision for a common way of describing people's needs and preferences.

Metadata, used in the research as an enabler, is only so in the ways required when it is interoperable, that is, is usable by many systems. Interoperability is an elusive quality. Interoperability does not include something as simple as two systems sharing the same software, as with an office software package, for instance. Interoperability across the vast complex systems of the World Wide Web (the Web) is demonstrated when systems that do not operate the same way, that are not multiple copies of the same software, that have not been purpose built to work together, share common standards of specification, and can share data. Metadata is data about data, and therefore itself data. Interoperabiity enables unprecedented quantities of data to be applied to a single instance and, as has come to be recognised in the context of the Web, 'quantity has a quality all its own' (often attributed to Joseph Stalin).

The research analyses and synthesises the metadata development work of the AccessForAll team and comes to the conclusion that there is a simple methodology for developing interoperable metadata standards, such as the AccessForAll standards. This process was not followed in the case of the first AccessForAll work with the result that all the problems reported in the research were encountered. The steps in the recommended process are:

The research distinguishes the context in which earlier accessibility work took place. It adopts a fiction and describes the development of the Web as happening in phases named Web 1, 2 and 3. In what might be thought of as a Web 1.0 environment, one-way publishing was the dominant activity. In the Web 2.0 environment, interactive user-participation publication happens across the Web in unpredictable ways, despite authors and publishers who provide well-structured, cohesive Web sites. In the emergent Web 3.0, data is organising itself. Many users 'Google' and they approach information from a range of perspectives and directions, often coming into resources through what is effectively a back door, and taking from resources what is of interest idiosyncratically to them, disregarding or discarding the rest.

The research also relies upon the interactivity and energy available from what is known as social networking that is well established within the Web 2.0 environment (Flickr, YouTube, LibraryThing, Facebook, etc). It exploits new technologies to solve an old problem and to share the responsibility for the problem well beyond the practices, knowledge, and expertise of the original resource authors. In this sense, it anticipates an even more active Web, where users design, create and control the published content and the content reorganises itself for the users. No longer do users just ‘browse’ a static, manually-published Web.

The research is not limited to classic 'Web pages', but includes access to all resources, including services, that are digitally addressed. AccessForAll metadata already describes digital resources and is being extended to describe a wider range of objects including events and places (ISO/IEC JTCI SC 36, 2008). Descriptions of the accessibility of those places and events will be Web addressable, so the necessary access to those descriptions will be 'on the Web'.

Background

The United Nations publishes a map (Figure 1) that shows involvement in the United Nations (UN) Convention for the Rights of People with Disabilities.

Figure 1: Map showing signatories of UN Convention for the Rights of People with Disabilities in November, 2008 (UN Enable, 2008c)

Australia signed the Convention in 2007 and ratified it in July 2008, with some declarations and reservations (UN Enable, 2008b), more than eighteen months after the Convention was adopted by the UN, Australians have been involved for many years in international efforts with a range of standards organisations (W3C, ISO, IMS GLC, CEN, and others), to ensure that information technology and digital resources are accessible to everyone. They have actively participated in the work of the World Wide Web Consortium [W3C] and others to curb the alienating effects of new multimedia technologies on the Web.

Making the Web accessible to everyone has proven more difficult than anticipated. While Roe (2007) considers the value of accessibility to be far-reaching, Constantine (2006) summarises the unfortunate reality; much as one might like to make the Web accessible, it is not accessible and is not likely to become so. The attitude of those who could make the difference is too often 'unhelpful'.

The same response has been experienced frequently by the author, who confronted some very influencial metadata experts to be told they just felt like cringing when they heard the word 'accessibility'. Not, they said, because they did not think it was important, but because they just did not know what to do and so felt bad. (The name of the Dublin Core proposed element was actually changed, for a while, to 'adaptability', for the benefit of such people (Nevile, 2004b).

People with disabilities who are alienated by inaccessibility are regarded by Australian law as discriminated against (HREOC, 2002). They are able to claim damages from those who discriminate against them if all relevant conditions are satisfied. This means Australia recognises a general right. It is incumbent on a victim to prove, within the legal system, that they have unreasonably suffered from discrimination. Although this has been done, reported cases are rare, as with other cases likely to provoke negative publicity. Such cases would normally be settled out of court where possible, and so not publicly reported. Such a legal situation does not operate as a major threat to large organisations, especially as the damages judicially awarded so far have not been substantial, e.g. Maguire v Sydney Organising Committee for the Olympic Games. (HREOC, 1999)

Accessibility efforts in many cases aim to make a single resource universally accessible to everyone. Universal accessibility involves providing the same resource in many forms so that people with disabilities can use the full range of perceptions to access it across all platforms, fixed and mobile, standard and adaptive. Universal accessibility is distinguished from individual accessibility or accessibility to an individual user. Many resources are individually accessible while not universally accessible and many universally accessible resources (as defined by the standards in use) are not accessible by some individual users (see Chapter 4).

Reinforcing the disinclination to worry about accessibility is the common belief that it costs a lot to make resources universally accessible (Steenhout, 2008). Frequently it is left to a semi-technical person in a relatively insignificant position within an organisation or operation to champion accessibility as best they can. Anecdotally, they frequently report that all was going well until the resource was about to be released. Then the marketing manager or some other more significant participant chose to add a particular feature and not be constrained by accessibility concerns. (In the 1990's, Nevile was responsible for the accessibility of two major government portals, the Victorian Better Health Channel and the Victorian Education Channel. In both cases, late requests for changes threatened the integrity of the sites but fortunately, in the end, the prior work made it possible to avoid any ill-effects of the changes).

Economic factors are important in the context of accessibility. Many assume that accessibility means more expenses when resources are being developed and more resources being supplied to the range of users of those resources. It is true that making an inaccessible resource accessible can take considerable effort, expertise and expense and, even then, is not always possible. On the other hand, some publishers are finding that by making accessibility a priority, they actually gain financially through cost savings (Jackson, 2004, Chapter 3).

Practicality is important. It has long been known that it is not always possible to make an inaccessible resource accessible without having to compromise some of the characteristics of the resource, depending on what sort of resource it is. If designers provide an attractive 'look and feel' for a Web site, for example, it may not be possible to have exactly that look and satisfy all the accessibility specifications. Additionally, those who are experts in accessibility are not usually designers but more often technical people. In practice, a designer who understands and works within the accessibility constraints is able to design creatively and avoid the accessibility pitfalls.

One common reason that resources are not accessible is that they are dependent on a software application that does not render the content, or does not control or display the content in ways that make it accessible to everyone. Many people with permanent disabilities use specialised equipment or software to gain access to content. Many users have mobile phones, and or use screens with content projected on to them, or printers, or old computers. Sometimes the content creator takes the end user into account. Unfortunately, this often means they arbitrarily anticipate too much, for example, that it will be printed on local-standard sized paper, in which case they fix the electronic version of the resource to match the way they expect it to appear on paper. This does not always work for the paper version because the local standards differ. Neither does it work for the digital version because rarely are screen sizes or windows appropriate as anticipated.

In cases where users have unusual needs or preferences, such as a need to change the font size or reverse the colours of the background and foreground, it is rare that the necessary changes can be made. It is possible, however, where the digital version of the fixed-print version is encoded correctly. The World Wide Web Consortium [W3C] has developed a technology that allows a single resource to be presented in a variety of ways, depending on the medium, and explicitly for the user to use a suitable form of presentation and override any made available by the publisher of the resource or the browser software [Cascading Style Sheets, CSS]. This is achieved by separating the content from instructions about how it is to be presented.

Many think of the Web as 'homepages' or Web sites. This is not sufficient. A Web page may contain links to documents that reside in databases, open or closed, and those 'documents' might be simply some application-free content, or they might be complex combinations of multimedia objects, even dynamically assembled for the individual user, and locked into specific applications. Even physical and ephemeral objects are included in the Web, such as when a cinema is listed as showing a film or a football match is advertised.

The Web Accessibility Initiative [WAI] is the arm of W3C that focuses on accessibility for the Web. WAI distinguishes between two classes of software used in this context; authoring tools and user agents. The classes include software that does very different things according to what it is being used to author or access, which can range from literature to computer code, images to tactile objects. Authoring tools should both produce accessible content and be accessible, according to the relevant WAI guidelines (Authoring Tools Accessibility Guidelines [ATAG]). User agents are the software applications used to access the content. They also should be both accessible and do the right thing with the content so that it is rendered in an accessible way (User Agent Accessibility Guidelines [UAAG]). (The user agents are often Web browsers but they can take many forms.)

The original WAI set of guidelines, one each for authoring tools, users agents and content [WCAG-1, ATAG, UAAG], have been in constant development or revision for more than a decade (Chapter 4). They have been adopted in various ways in many countries and used by developers all around the world. Despite this incredible effort, the Web is still far from accessible to everyone (Chapters 3, 4). The underlying principle for these guidelines has continued to be universal design, to be achieved by having a single resource that can be used by everyone.

In recent years, total dependence on the WAI work and its derivatives (such as s. 508 that was added to the US Disabilities Discrimination Act [DDA]) has been re-examined and a range of post-production solutions are being proposed. In particular, methods have been developed that support increasing the accessibility of a resource by a third party, unrelated or connected to the original publisher. ubAccess, for example, developed a service [SWAP] that could assist people with dyslexia who were having problems with resources, without reference to the original creator of the resource. In a similar way, a service called Access Monkey operates to assist blind users without reference to the original author of the resource (Bigham & Ladner, 2007).

More and more such services are emerging. What is significant is not simply their number. It is that they represent a significant shift in thinking about accessibility. If resources are not going to be created universally accessible, or found in a universally accessible form, and it is unlikely there will be a significant change in this situation, it makes sense to think more about what can be done post-production.

Post-production techniques were a feature of the 2007 OZeWAI Conference. Pierre Frederiksen demonstrated how to automatically make a complex table accessible post-production to users not relying on vision and Charles McCathieNevile showed how an established inaccessible technology can become an accessible technology simply by the adoption of suitable encoding techniques (OZeWAI 2007). McCathieNevile demonstrated the techniques for Asynchronous JavaScript and XML [AJAX]. Hudson and Weakley (2007) argued that now social networks are common on the Web, collaborative action can be taken quickly. They argued that the emergence of techniques for developing or repairing inaccessible technologies, as in the case of AJAX, is quick when developers and users and others involved all agree on a goal and the effort is shared throughout the community. Such adaptations of not-yet-accessible resources or resource components, post-production, offer great hope in the field of accessibility research.

Going a little further, the FLUID project is developing interchangeable user interface components that can interpret and present content in ways that are accessible to individual users (2007). This depends on content being made so it is not application or interface specific, but free to be adopted and adapted by any standards conformant applications, interfaces, and thus accessible to all who use it.

An outdated view of accessibility and of the Web

The original use of the World Wide Web was to enable a few people scattered around the world to work together on shared files located on their own computers. It made the files discoverable using a Uniform/Universal Resource Identifier [URI] so they could be accessed using the HyperText Transfer Protocol [HTTP]. The goal was collaboration among a few scientists. In the first decade of widespread use of the Web as an information and communication technology, the main activity was the publication of resources. This involved the use of HTML encoded files that offered embedded links, embedded multimedia resources and may have had cascading stylesheets [HTML 4.01]. The publishers relied on third party HTTP or Web servers to deliver the files to users who browsed the publications. Now, as is recognised by the new name 'Web 2.0' (see below), all sorts of interactive, collaborative and shared activities are being undertaken using a wide range of technologies. Already we are seeing extensive publications that are user-driven, and the publishers are often responsible for creating an environment in which the users contribute content to be published (e.g. the image publishing site Flickr).

The research establishes that the dominant model of accessibility work is still grounded in the early populous Web, a network of static documents that may be updated but are usually from a single source. In this thesis, the term Web 1.0 is used to designate this Web as it was in its first decade, 1995-2005. O'Reilly (2005) used version terminology to differentiate between the uses of the Web and to draw attention to more recent developments in the way people use the Web. Of course, it should be noted that the Web does not, in fact, have versions (O'Reilly, 2005) and this terminology is more about how it is used than what it can do.

Web 1.0 work assumes editorial control over publishing, even where the authors come from a single organisation and this task is undertaken by a number of people. In such cases, in fact, many organisations impose both style guides (or the equivalent) on the authors and/or provide templates within which those authors have constrained scope for their content. In such circumstances, it might be possible to force adherence to certain style standards, as it was in the earlier days when documents to be printed were encoded in Standard Generalized Markup Language [SGML] (the predecessor of HTML). The model also assumes that users of Web resources will interact with them as their author intended. More and more this is proving not to be the case as people use search engines, dynamic feeds from within Web sites, etc.

Many people still do not recognise that they can use standard Web pages and Web authoring tools in almost exactly the same way as they use non-standard proprietary office tools. Web authoring tools have many of the same features as word processors, including facilities to format, print, exchange and manage other documents. Many people are still using proprietary-format office tools that do not offer the accessibility that is now possible.

Organisations in which proprietary office tools are used form sub-cultures around those tools, and participants develop materials (resources) that suit the particular local software tools. They are often not aware that their single resources could be as easily created and managed but far more flexible and interoperable not only between software systems, but also across ranges of modalities (on paper, on individual screens, as presentations on large screens, read aloud, etc.). Proprietary interests and competition have encouraged proprietary developers to distinguish their software by adding features often regardless of the inaccessibility simultaneously introduced by those features.

At the time of writing, there is a worldwide concern about the adoption of the Microsoft specification Office Open XML as an international standard for documents. One reason for the concern is the problem of accessibility that may flow from such a decision (Krempl, 2008). Portable Document Format [PDF], another proprietary format, has long proved a problem for accessibility and continues to do so despite being an ISO standard (W3C PDF, 2001) and despite the possibility of using it to produce accessible content.

The research establishes that the historic view of accessibility is no longer effective.

The complexity of satisfying the original guidelines is shown to be out of the range of most developers. There are too many techniques involved; they are not explicit; they cannot always be tested with certainty; they do not completely cover even chosen use cases and are not intended to cover all user requirements; they are contradictory in some cases; they have not been applied systematically, and anyway, they do not apply to all potential information and communications. All of these claims are documented in this thesis.

In addition, it is not appropriate to be concerned only with how to give access to Web 1.0 type activities. All users need to have access to all the facilities of the evolving Web, including contributing to its evolution.

A new approach to accessibility for Web 3.0

This thesis does not claim to be alone in making the claims above: there are many researchers working on similar problems, as shown in the thesis. Their work is considered in detail in the research.

What this thesis offers is an argument in favour of an on-going process approach to accessibility of resources that supports continuous improvement of any given resource, not necessarily by the author of the resource, and not necessarily by design or with knowledge of the original resource, and often by contributors who may be distributed globally.

In taking this stand, the research anticipates the evolving Web. It argues that the current dependence on production guidelines and post-production evaluation of resources as either universally accessible or otherwise, does not adequately provide for either the accessibility necessary for individuals or the continuous, evolutionary approach already possible within what is defined as a Web 2.0 environment. The research assumes a distributed, social-networking view of the Web as interactive, combined with a social model of disability, for the management tool using machine-readable, interoperable AccessForAll metadata. It goes further and anticipates the ability of data itself to form relationships and logical connections within the Web. It argues this can support continuous improvement of the accessibility of the Web with less dependence on the part of original content developers and better results for individual users.

As outlined above, there are a number of ways to make resources accessible. Relying solely on authors to 'do the right thing' by following the universal accessibility approach has generally failed to make resources universally accessible (Chapter 4) but many resources are nevertheless suitable for individual users, if only they can find them. Similarly, most resources that are universally accessible are not discoverable as such.

In Europe, there have been moves to apply metadata (labels) to resources, to catalogue them, and declare their accessibility in terms of conformance with various available specifications: the UK government has mandated certain provisions (BSI,2006; Sloan, 2005; Appendix 6) and the European Centre for Standards (CEN) supported a project (later abandoned) led by EuroAccessibility for an accessibility conformance mark for use in all European countries (RNIB, 2003).

There have also been reservations expressed about the proposed 'conformity labeling' approach (Phipps et al, 2005). The current research also challenges the wisdom of that practice. As there are often legal implications for having resources that are not accessible, even if there is not an economic incentive that might bias evaluations, it is hard to know which evaluations to trust. It is also very hard to evaluate accessibility accurately. One reason for the problem with the evaluation of accessibility is that only some of the criteria can be tested against absolute standards, as most depend upon human judgment. This causes problems because many people can manage to do the technical tests using automatic evaluators. They do not realise they also have to do the human-based user testing, and when they do, they lack the knowledge, resources and expertise to do this properly. In fact, to rectify this situation, those developing specifications, such as the World Wide Web Consortium's Web Accessibility Initiative, are endeavouring to make all specifications testable against absolute values. Unfortunately, to achieve this, they appear to be compromising some of the specifications (Hudson & Weakley, 2007) and end up having to ignore the needs of important communities of users such as those with cognitive disabilities (Moss, 2006; [WCAG-2]).

Metadata (labels) that merely identify resources that have been marked as accessible are not particularly reliable and anyway, as is shown in Chapter 4, conformance with the best-known guidelines does not necessarily mean a resource is accessible to the individual who wants it. Certainly, such metadata does not say if the resource is optimised for any particular individual user seeking it. More specific metadata is required if it is to be useful to the individual user. This has been recognised by the authors of the WCAG guidelines and there is provision in the revised version of WCAG [WCAG-2] for metadata as a result of the AccessForAll work .

If resources are to be made more accessible post-production, their inadequacies will need to be discoverable prior to delivery. When found to be inaccessible, any missing or supplementary components, or services to adapt them, will also need to be discoverable. Resource descriptions, like catalogue records, can usefully contain descriptions of the accessibility characteristics of resources without any need for declaring if the resource is or is not conformant to a comprehensive standard. Such characteristics' descriptions are known as AccessForAll metadata and discussed in detail in Chapter 7. AccessForAll metadata has been adopted by four major standards bodies. First, the IMS Global Learning Consortium [IMS GLC] for the education sector. Then the Joint Technical Committee of the International Organisation for Standardization/International Electrotechnical Commission. Its, Sub-Committee 36 [ISO/IEC JTC1 SC36], endorsed and adopted it again for the education sector. The Dublin Core Metadata Initiative [DCMI] is adopting it for general metadata, for all sectors, and Standards Australia has adopted if for the AGLS Metadata Standard [AGLS], for all Australian resources.

In addition to metadata that describes the accessibility characteristics of resources, it is necessary to define metadata to describe the functional needs and preferences of users.

'AccessForAll' metadata is best used to match resources to users' needs and preferences, automatically where possible. Determining how such a match might be achieved in a distributed environment is a continuing interest of the author and especially colleagues at the Research Center for Knowledge Communities [KCRC] in Japan, in as much as it relates to the use of the Functional Requirements for Bibliographic Records [FRBR], OpenURI (Hammond & Van de Sompel, 2003), and possibly GLIMIRs (Weibel, 2008a). This highlights the significance of the metadata as defined, the potential matches, and the ways in which AccessForAll metadata contributes to the accessibility process.

'Usability' is well established as a criterion for the utility of a resource (Nielsen, 2008). A flexible approach including usability in a loose sort of 'tangram' model could significantly improve the Web's accessibility (Kelly et al, 2006, Kelly et al, 2008). The AccessForAll metadata enables the management of resources in such a process with adaptability for personal needs and preferences for a better result.

This thesis describes the background, theories, design and development of the metadata framework, as implemented and documented in the various published or forthcoming standards, and work associated with its adoption by various stakeholders. In doing so, it exposes the problems of designing metadata frameworks for interoperability and finally focuses on the need for a tightly specified framework for such metadata if it is to achieve the goals set for it by the context. The thesis thus starts with the focus on accessibility and later moves it to metadata. The proposed role for the metadata is the managment of data relating to accessibility, so both are significant in the research.

Understanding and significance of accessibility

Understanding accessibility is not easy given the huge number of different contexts and requirements possible. In addition, there are many definitions.

For the purposes of the research, accessibility is defined as a successful matching of information and communications to an individual user's needs and preferences to enable that user to interact with and perceive the intellectual content of the information or communications. This includes being able to use whatever assistive technologies or devices are reasonably involved in the situation and that conform to suitably chosen standards. Explanations of the more detailed characteristics of accessibility are considered in Chapter 3.

Throughout this thesis there is reference to literature that reveals two things: a current common approach to accessibility that is significantly reliant on universal accessibility, as promoted by the World Wide Web Consortium [W3C], and a significant failure of that approach to make a sufficient difference despite the need for accessibility.

In 2008, despite the introduction of quite stringent provisions regarding the accessibility of government sites, SiteMorse (2008) report that only 11.3% of UK government websites surveyed passed the WCAG AA test mandated for such sites (Cabinet Office, 2008). (The sites were tested only with automated tests, so the results are only indicative of 'universal accessibility'.) Those with needs in terms of access in Europe are estimated to include 10-15% of the population who have permanent disabilities and the number is increasing as the population ages (European Commission, 2007).

Microsoft Corporation commissioned research that suggests the benefits of accessibility will be enjoyed by 64% of all Web users (Forrester Inc., 2004). In 2004, the United Kingdom's Disability Rights Commission [DRC] reported on the accessibility of 1,000 UK Web sites (DRC, 2004). They showed that 81% of Web sites failed to meet even minimum standards for Web access for people with disabilities. Later, at a press conference, the DRC claimed that even sites considered prima facie to be demonstrating good practice, in fact failed to satisfy minimum standards when fully tested by the DRC. These reports have been endorsed more recently by the United Nations' Global Audit of Web Accessibility (Nomensa, 2006).

AccessForAll philosophy

The more information is mapped and rendered discoverable, not only by subject but also by accessibility criteria, the more easily and frequently inaccessible information for the individual user can be replaced or augmented by information that is accessible to them. This, in turn, means less damage when an individual author or publisher of information fails to make their information accessible. This is important because, as is shown ( Chapters 2, 5), making resources universally accessible is burdensome, unlikely to happen, and does not guarantee that the information presented will, in fact, be accessible to a particular individual user. It also means that distributed resources need to be managed so they can be augmented or reformed by components that are not originally a part of them or not intended to be associated with them. This can be facilitated by suitable metadata.

Widespread-mapping of information depends upon the interoperability of individual mappings or, in another dimension, the potential for discovering distributed information maps in a single search. The ancient technique of creating atlases from a collection of maps is exemplary in this sense (Ashdowne et al, 2000). Being able to relate a location on one map to the same location on another map is achieved easily when latitude and longitude are represented in a common way, or when the name of one location is either represented in a common way, such as both in a certain language, or able to be related via a thesaurus or the equivalent.

Atlases would not be useful if every map were developed according to different forms of representation; the standardisation of representations enables the accumulation of maps to form the global atlas. In the same way, the widespread mapping of accessible resources on the Web is achieved by the use of a common form of representation so that searches can be performed across collections of resources. Interoperability is typically said to function at three levels: structure, syntax and semantics (Weibel, 1997). Nevile & Mason (2005) argue that it does not operate at all unless there is also system-wide adoption (see Chapter 11). System-wide adoption might now be thought of as substantial adoption.

The original AccessForAll team (the AfA team) at the University of Toronto worked to exploit the use of database classifications in the discovery and publication of their digital content in a way that could increase accessibility for their university users [TILE]. They exploited a common way of describing the resources and resource components so they could mix and match them to user’s individual needs and preferences. When this activity widened in scope, as IMS GLC work [IMS Accessibility], there was a need to think about how the necessary descriptions could be shared across institutions. The aim was to develop specifications for the description of resources so these could be shared. The hope was to specify a common vocabulary for describing the resources. It would specify a ‘machine-readable’ form for those descriptions, defining a set of terms that could be used and how they might be used by a computer.

Further work led to yet another set of specifications, now an international standard developed by many national bodies and experts [ISO/IEC 24751-1:2008; ISO/IEC 24751-1:2008; ISO/IEC 24751-1:2008]. The aim has not changed. It is that individual users everywhere can find something that will serve their purposes in a way that is independent of their choice of device, location, language and representational form.

The research highlights the need for interoperability of metadata specified in those standards. Unfortunately, it shows that so far this has not been attained as it might be. On the other hand, developments within the relevant standards organisations have also recognised the problems of interoperability, and on-going work to which the author is contributing is engaging with the problems. Specifications for metadata for learning resources have been under development for some time but by late 2008 it has become clear that a number of national bodies and experts are very concerned that metadata is developed only in conformance with a deeply interoperable model, and this seems to mean within a Semantic Web Resource Description Framework (W3C RDF, 2008).

The AfA work takes advantage of the growing number of situations for which metadata is the management tool for digital objects and services and information about people's needs and preferences with respect to them, so that well-described resources that are suitable can be discovered. AfA philosophy includes, in addition, that resources should be able to be decomposed and re-formed using metadata to make them accessible to users with a range of devices, locations, languages and representation needs and preferences. Chapter 11 expands on some significant if not yet widespread adoption of this idea. AfA metadata can be used immediately to manage resources within a shared, closed environment such as the original one established at the University of Toronto where the AccessForAll approach was conceived.

This thesis argues that there is greater potential for benefit from accessibility metadata if it is not only expressed in a common language but within a common description framework. That is, rather than just sharing particular ways of describing particular characteristics of resources, it should be possible to describe those characteristics in a common way, so that all the descriptive sets are interoperable. This is considered the essence of good metadata and now also being advocated by some as the appropriate base for all ISO/IEC JTC1 SC36 metadata work (ISO/IEC CD 19788-1).

A metadata approach

In the case of the AccessForAll projects, and others, Nevile is a metadata scientist. Her research is about how a metadata framework should be designed so that metadata developers can share their metadata and rely on its interoperability at a machine level. The research is grounded in work operating at a different level, involving the definition of metadata schemata (herein called schemas).

Using a common term such a ‘dog’ is not problematic when talking to a child and differentiating between a horse and a dog (or wolf). It is a serious problem if it is used indiscriminately when it comes to the classification of species.

As explained in Chapter 6, the value of metadata usually is in its interoperability, and this depends, like the definition of ‘dog’, on broad and deep interoperability with other classifications. Local utility does not make such demands, and much has been achieved in the past with what is now often called metadata, the classifications within databases of content (what are known as fields in database terminology). In the Web 1.0, database-driven customization of Web content is very useful, especially where there are fast applications, connections, and even shared (common) fields so queries can be distributed. The vast library ‘union’ catalogues work this way.

A characteristic of Web 2.0 is that so-called users contribute to the Web. They do not just consume content but do such things as contribute content and comment on content, theirs and others, for example by adding tags to describe it. In many cases, the users of Web 2.0 are filling in the missing gaps in databases designed by the content publishers, who now publish but do not always author content.

Web 3.0 is expected to be upon us very soon. There is speculation about what it will be but concensus is firming up around what is known currently as the 'Semantic Web'. Kevin Kelly (2007) envisions yet later versions of the Web as an environment in which people and objects have become part of the Web, it has become 'one', with data playing an 'intelligent' role.

Increasingly, what we might call Web 3.0 activities are emerging that involve structuring of content in ways not anticipated by the original authors or publishers. For example, imagine a site that publishes photos submitted by users. Imagine a second site that refers to content on the first site and provides text descriptions of the photographs. In this case, the second site provides some original content (perhaps contributed by the general public) and ways of structuring distributed content without any need for agreement about the form of the content with the original authors. The second site developers do not republish the photographs. They might be using metadata about the original content, however. If the second site developers want to provide a service for users, and offer text descriptions of photographs from wherever, they will want to be able to use metadata from wherever. They will use the metadata to determine things like the date of the photograph, the photographer, the camera used and other things that they will want to be able to use in their management, or include in their description, of the photograph.

Imagine a site where students who cannot access some multimedia learning materials provided by their institution about Hamlet can find other resources that deal with the same topic but in text. They will want computers to be able to know what work they are learning about, in this case Hamlet, and to find other materials about the same work. In addition, they will want to know in what format the alternative materials are available. There will be a need for access to information, descriptions of the resources, if users or their agents are to match the alternatives to the originals and also, to be made aware of the existence of the alternatives when they discover the originals. This is so even if the creator of the originals does not know the alternatives exist, and especially if they do not point to them.

The interoperability of metadata depends on more than the language used, as explained in Chapters 6 and 11.

Metadata research is necessary to determine how metadata should be specified. It is not about the metadata itself – that is implementation, and the work of metadata engineers. The AccessForAll work started at the level of metadata implementation [TILE]. Through work on that implementation, development as it is often called, research questions have emerged that have led to a scientific view of metadata.

At the Metadata Research Center, School of Information and Library Science, University of North Carolina at Chapel Hill, a number of projects for developing metadata for specific domains have been funded and undertaken as research [MRC UNC]. A typical example is provided by the KEE-MP project:

Such research does not depend upon standard research techniques (see Chapter 2), but nor is it development in the usual sense. While the direct output may be a prototype product, the research is about metadata, a branch of information science. Some of what is learned is inevitably what is not supported by metadata as it is used, and how effective the evolving principles are, and what could improve them. The KEE-MP project involved expertise about films. The current research is based on the effectiveness of the evolving principles of technical accessibility development and ways to improve them.

AccessForAll metadata research and development

In the current research, the basic computer science task of classification in first normal form (IBM, 2005), in a functionally unambiguous way as used in databases, is abstracted into the field of metadata.

At one level, the focus is on how to classify the objects of concern. At the next level, it is how to define and structure those classifications, the semantics of those classifications, what terms are to be used. Neither of these levels guarantee interoperability, or even depend on it. They are concerned with establishing what can be thought of as a vocabulary for describing the objects. Such a vocabulary can be shared, making it a common vocabulary. Metadata research is concerned with how to make it an interoperable vocabulary, and that depends on its syntax and structure being shared.

Implementers and developers work to unambiguously classify objects building databases and thesauri. The science of metadata, how to express and make interoperable such classifications, evolved from the librarian's discipline of cataloguing, inheriting many principles but explicitly rejecting others or adapting them, and adding some new ones. The role of technology, and hence the syntax and structure of the classifications, is significant in metadata work alongside the semantics that are always in focus in library work. Metadata research is an evolving field, given the changing nature of the technology and its use, especially the technology and use of the Web.

The current research also concerns the effectiveness of the evolving principles of technical accessibility development and ways to improve it. It was undertaken within a context of collaborative work on accessibility for a number of organisations involving a number of players over an extended period (Figure 2).

Metadata research projects often involve a multi-disciplinary team including both developers and researchers. In as much as the research requires the use of new technologies, and they need to be built and tested, developers are often essential to the work. It is also important to have people to design the system, understanding the requirements for any metadata to be used. There is finally, of course, a need for subject experts, who contribute information about the structure of the knowledge of the domain, and how it is used.

The Assistive Technology Resource Center [ATRC] at the University of Toronto has a proud record of research and development. In the field of accessibility, they have significant achievements and, specifically, were leaders in the use of database technologies to adapt resources to users’ individual needs, with their product ‘The Inclusive Learning Exchange’ [TILE].

TILE is a database application in which particular inputs prompt certain responses from a computational system.

In the AccessForAll interdisciplinary development team, there have been seven major players: Jutta Treviranus, Anastasia Cheetham and David Weinberg, in particular, from the Assistive Technology Resource Center [ATRC] at the University of Toronto, Canada (University of Toronto, Canada); ; Madeleine Rothberg from WGBH National Center for Accessible Media in Boston, USA (WGBH/NCAM); Liddy Nevile from La Trobe University, Australia; and Andy Heath from the University of Sheffield (now at the Open University) and Martyn Cooper from the Open University, United Kingdom (Open University, UK).

All in the team have been involved in accessibility work for a number of years but from different perspectives. Nevile is the metadata scientist/researcher in the team, Cheetham and Weinberg are responsible for the development of the prototype TILE, Heath is an expert in programming, and Rothberg, Treviranus and Cooper are responsible for major accessibility projects in education. Treviranus is the outstanding accessibility expert. Treviranus is the Director of the ATRC and Chair of the W3C Authoring Tools Accessibility Guidelines Working Group [ATAG WG], among other things.

The AccessForAll work has been undertaken in a number of contexts (as explained below) but always with the core team leading the efforts. The group emerged from the work being undertaken by the IMS Global Learning Consortium [IMS GLC] when they adopted the ATRC model, and has moved to other contexts, as explained below. Nevile, the Chair of the DCMI Accessibility Working Group (now the Accessibility Community), is responsible for AccessForAll finding its way into the DCMI world of metadata and has been responsible for developing the Accessibility Application Profile (or Module) for DCMI and all the schema and documentation required for an international technical standard (DCMI Access).

The local aim of the metadata research is to find a way to enable the AccessForAll approach in a variety of formats with the greatest possible potential for interoperability between those formats. As always, those leading in AccessForAll work are involved in many overlapping and, at times, conflicting communities (Figure 3, 4). Consequently, this work has been undertaken in a socio-technical more than a scientific environment.

This thesis argues that metadata is an enabling technology for a shift to an AccessForAll approach to accessibility. It is at the core of the research as is accessibility itself. From the beginning, Nevile's involvement has been based on questions that have arisen in the Dublin Core Metadata Initiative context, motivated by earlier work in both the metadata and accessibility fields, and focused on the potential role for metadata in increasing accessibility. The need to distinguish between compliance and functional metadata is a good example. It was the subject of discussions at an open meeting in Seattle in 2003 (Nevile, 2003d).

There is yet another dimension to the research that should be noted. This is the design dimension.

Discovery of resources always involves some sort of matching, even for avoidance purposes. The design of the metadata to be matched depends, usually, on the requirements in the circumstances. In the case of the original prototype for TILE, for example, the requirements would have been ascertained and then the metadata would have been designed to provide the necessary information. In order to match a resource to a person who cannot use sight, it is necessary to determine if a resource requires sight as a perceptive mode. If so, it might need to be avoided. It will be necessary to have a way of determining if the resource does need to be seen and recording this information so the resource sorting process can be applied when it is called upon.

TILE was built to provide users with the best match of digital resources from a given set of resources, all authored and available in-house. The AccessForAll approach to accessibility, generalizing this idea from TILE, aims to operate in a global context. As soon as the first set of terms were developed, members of the team felt pressure to work on other types of objects, including in the mix objects such as people who might be personal assistants to users with disabilities. Where there is a human assistant, the functional requirements of the user may not change but there is a new set of skills to be configured into the matching process. Similarly, when some of the resources are non-digital, their characteristics will not be described well by using a profile for digital resources. For example, digital text is very different from printed text.

Another way of thinking of the difference between the work of the original TILE developers and the metadata research is to think of what is produced by the two groups. As a result of the work of the TILE developers, resources within the TILE system could be delivered to users in a chosen way. As a result of the work of the metadata research, it should be easier to write metadata to support AccessForAll accessibility in new circumstances. The metadata framework, alone, does not do anything. The product of the research is a comprehensive explanation of how to do something, and why it works or otherwise, not the doing of it.

As stated above, the metadata researcher’s task is not to implement or even design a set of terms for use in the circumstances, but rather to design a framework in which such sets can be designed by others, with confidence that they will all be interoperable. In Chapter 10, it is explained how metadata is mixed and matched, in other words, how it interoperates in useful ways. Here, suffice to say that this dimension of the research that is about enabling the design of metadata, what might be called the meta-design of metadata (Fischer & Giaccardi, 2006), is central to the research.

In addition, the research is transdisciplinary, and so requires appropriate competencies. As Derry & Fischer (2007) point out,

Derry and Fischer (2007) point to the problem of who owns work developed in a socio-technical environment. They report that this type of work is a priority in post-graduate research because it is in the ‘real’ world. But they also assert that, so far, there is no clear research about how to evaluate this type of work. It may have been considered a problem, for this thesis, to make explicit what is the contribution of the research it reports. Rather than shy away from the collaborative nature of the work undertaken as part of the whole AccessForAll endeavour, however, this thesis aims to celebrate that collaboration. It asserts it is a strength of the work that it is based on actual practice in a transdisciplinary, socio-technical context.

Research objectives

The focus of the research is how metadata can provide the infrastructure for AccessForAll practices to make the Web more accessible.

With respect to metadata, the research challenges the structure, the syntax and the semantics of the AccessForAll work. It includes:

The AccessForAll standards, and other products of the research, are published elsewhere and, increasingly, implemented and further researched.

With respect to accessibility, based on estimates of the current accessibility of the Web, the research challenges the theoretical foundations of previous work. It adopts a new base to support inclusion and the UN Convention for the Rights of People with Disabilities (UN, 2006). It includes:

Research method

The first phase of the research involved coming to terms with the requirements of the Web Content Accessibility Guidelines [WCAG-1]. The author had worked with the Authoring Tools Accessibiity Guidelines [ATAG] editors in the hope that by specifying how authoring tools should work, the guidelines would lead to the use of tools that enabled authors to make increasingly accessible resources without having to think about accessibility. The reality was that the major tools developers did not adopt the specifications. The author tried to build a site that would provide a functional approach to accessibility for authors who were trying to achieve it. This site took a different approach from the WCAG but with the aim of implementing it. The goal was to have a quick-lookup menu so that a resource author could look directly for what they needed - how to make symbolic mathematics accessible, for example. The author hoped that the site would become an interactive site, to which others would contribute.

Appendix ??? contains the site as developed. It did not achieve its ultimate goal but was used for several years by La Trobe University as their guidelines for accessibility and significant parts were adopted in the "IMS Guidelines for Developing Accessible Learning Applications" (Barstow & Rothberg, 2002).

The next phase involved the development of metadata for the Dublin Core Metadata Initiative through its Accessibility Working Group. This work showed that existing DC terms were inadequate for accessibility descriptions and pointed to the need for some special metadata but did not make significant progress until the early AccessForAll ideas were introduced. Meanwhile, the author was working with the INCITS V2 team to understand what metadata would be required for the alternative interface access protocol for the universal remote console (Sheppard et al, 2004).

In 2003 the author brought together the major players from the various accessibility working groups with the aim of making a collaborative effort for the development of accessibiity metadata. Judy Brewer from W3C, Jutta Treviranus from the University of Toronto, and others met and discussed how metadata might be involved in the effort to make the Web more accessible (DC Accessibility Meeting, 2003).

The author was working with the IMS AccessForAll team specifying how the accessibility characteristics of resources and the accessibility needs and preferencs of users, might be described so they could be matched. This work depended enormously on the substantial 'accessibility' input from the University of Toronto experts but the author was conviced its success would also depend upon good metadata principles. There wer many vigorous and some heated discussions about this in weekly teleconference meetings and a number of multiple-day face-to-face working sessions. The author was struggling to articulate the principles, as it happens concurrently with the DCMI's own struggles, and often thought to be infuriatingly pedantic about these principles. The problem was that TILE was an operational system and so, some thought, proof of what was required, while the author was convinced that effectiveness and longevity of the work was dependent on metadata principles.

It was of enormous good fortune that those involved in the work were, in fact, passionate about their work and the result of many hours of debate usually satisfied all concerned and moved the work closer to what the author considered better metadata. It is most likely that this would not have happened had there not been such a competent team and so much engagement with the issues. It is probably not insignificant that most of those involved were used to development within an academic research environment.

The research was not bounded by the AccessForAll context. That work was taking place in the educational metadata milieu while the author was concerned to publish it within the wider DC general metadata world. This meant the author had to work on the DCMI approach to metadata and their principles. As said above, these were still in development, and so there was a need for engagement with that work. At one point the author contributed to the debate about the expression of the DCMI's metadata model, which led to its further refinement (Pulis & Nevile, 2006).

Finally, the ISO JTC1 SC 36 that is concerned with standards for Learning, Education and Training, at the time of writing is developing a standard for metadata. The major standard for educational metadata, Learning Object Metadata [LOM] was developed many years ago, and proved cumbersome and has not been implemented easily. When developing metadata in 2000 for the Victorian Government's Education Channel, the author chose to use modified DC metadata instead of the LOM. This led to new terms for DC metadata and the first DC application profile. The DCMI formed an education community and they are active in developing metadata for education that complies with the Dublin Core Abstract Model [DCMI DCAM]. In 2001, the author initiated a meeting between the DCMI and the LOM communities. This led to what is known as the 'Ottawa Communique', an agreement that these communities would work together to achieve greater interoperability between the two sets of metadata. A joint paper was published in 2002 that explained the similarities and problems (Weibel, 2002). All of this is relevant because the ISO work on a new Metadata for Learning Resources standard is required to match the DCAM and be as compatible as possible with the LOM in doing so (ISO/IEC JTC1 SC36 resolutions, 2007). Not only does this work provide a further challenge for the author and the research but it is also able to draw upon some of the research (particularly the work in Chapter 11).

Chapter Summary

In this Preamble, the scene has been set for the substantive work that follows. The development of a new way of working on the problem of accessibility has been introduced as not just a response to the lack of real success with previous methods, but also a response to the changing technological context in which this work takes place. Metadata research has emerged as a substantial field in the last ten years and metadata development has led to its adoption for resource management within digital systems. In addition, earlier understanding of disability according to a deficit model has been replaced by a social inclusion model that avoids distinctions between people with physical or other medical disabilities and the general public, assuming that everyone, at times, is disabled either by their circumstances or by temporary or permanent human impairment.

In the next chapter, many of the components considered in the research are defined in greater detail and the research is further described.

Chapter 1: Preamble

Introduction