Digital Content Creation and Navigation: Interoperability and Accessibility
Report to SHM Productions Ltd
Date: 15/10/2001
Article
0. Background
0.1 All the background information connected with this Report is at Nottingham City net - Site Accessibility Audit: a Brief for humanITy. www.nottinghamcitynet.co.uk
1. Introduction: Characteristics of Contemporary Digital Information Systems
1.1 Advances in such technologies as scanning/OCR and digital photography tend to work against the labelling metadata in different ways, such as "heading", "body text" and "graphic" and simply present screen-loads of data as single graphics. At best these can only be adjusted/customised in bloc so that, for example, if a text element is magnified the background is magnified to precisely the same degree.
1.2. Most systems interrogation and interaction depend upon the use of a qwerty keyboard and/or mouse. This may be optimal for a wide variety of people and tasks but the most widely used information navigation and implementation device in use is the numeric keypad, both for television remote control and for telephone use.
1.3 Information systems designers, to the extent that they think about it at all, assume a high degree of literacy and ability to interpret the meaning of graphics.
1.4 Links to translation engines, précis engines and text simplifiers are rare.
1.5. Apart from "Help", tools are rarely integrated into information systems.
1.6. Although we refer to the Internet, systems usually assume that traffic is one way, from a product to a consumer, or at best two-way between a producer and a consumer. The quintessential benefit of the Internet, which makes it qualitatively different from other communications channels, is that it can facilitate multiple, asynchronous creation and transaction. Systems rarely facilitate this.
1.7 In spite of its potential for providing elements of data with a variety of attributes and search routes, hypertext expressed in HTML is generally thought of simply as a mark-up language, similar to SGML but simpler. It is used simply to allow a browser to render information in 'flat form' re-uniting the content with style characteristics. This 'flat form' presentation tends to be reflected in 'tree' structures for searching which allocate a single attribute to data elements.
1.8 Where content adjustability is available this is usually confined to data; but if metadata such as menus are inaccessible, the accessible data cannot be reached.
1.9 HTML is frequently 'bent' to provide new features: the greater the 'bend' the less the interoperability.
2. General Comments
2.1 All of these shortcomings militate against optimal, customised access, where "optimal" refers to the cost/benefit to the supplier and the cost/benefit to the user. Calculating the cost/benefit to the producer is difficult but it differs from the same calculation in the production of analogue material because digital publishing is at once provisional but forever.
- Investment now will generate a better chance of longer, down-stream use
- The cost of making changes is low
- Two sides in a current debate can be summarised as follows:
- Providing increased accessibility robs products of their competitive advantage because it delays their launch
- Accessibility guarantees a bigger long-term market
2.2 In purely economic terms, this debate depends on balancing two factors:
- Are we concerned with the time between conception and launch or the time between conception and profit?
- Are we interested in a short-term profit or a long-term income stream?
2.3 In considering these questions it is important to understand the demographics of digital access. These are partly based on the 'classic' socio-economic divide but there are other important factors, which partly overlap but partly extend beyond that traditional divide; these are:
- Literacy/cognition
- Language
- Ageing
- Gender
- Disability.
It can easily be argued that people with literacy/cognitive problems and language difficulties and people who are elderly, disabled or women constitute a large part of the socially advantaged group on the 'wrong side' of the 'digital divide' but the current paradigm is both too rigid and too stark. Many people who fall into the above categories have average and above average incomes and the 'divide' narrows as you move away from PC-based technologies towards consumer electronics mediated access and the telephone.
2.4 European and domestic legislation, regulation and conditions of licences also have to be taken into account in framing an optimal information access policy.
2.5 From the user's perspective, time may be as important a resource as money. There is strong evidence that the economics of pc-based connectivity are a barrier to participation but, for example, in the area of e-commerce, sales are being lost because of the difficulty in completing transactions. In the age of unscheduled television there will be millions of programmes on offer but if an individual spends half an hour finding what he wants to buy that is half an hour spent not viewing purchased product. The key is customisation. In the short-term, customisation will be a relatively tedious process involving diagnostics and parameter setting but as systems become more sophisticated the process will become increasingly heuristic. Already it is possible within NT to establish a set of individual profile settings and migrate these within a system. In time these will either be stored on a smart card or a personal web site so that they can be 'called' by the user at any point of transaction. The next step is the use of heuristics, which are of enormous economic benefit in the area of on-demand digital entertainment (particularly as television becomes unscheduled).
3. Solutions
The following sub-Sections correspond with those in Section 1., e.g. 3.1 refers to 1.1 above:
3.1 Data Stratification and Labelling
3.2 The Power of Nine
3.3 Discrete and multi-media access
3.4 Language engineering and simplification engines/tools
3.5 Integration of procedure and processing tools into systems
3.6 All-to-all processing systems and collaborative process and procedure tools.
3.7 Multiple label and multiple attribute allocation to data
3.8 Uniform Customisation
3.9 XML with Style Sheets
3.1 Data Stratification and Labelling
Think of the contrast between the content on a compact disc and a photograph. Ever since the development of electronic recording, engineers have tried to divide recorded sound into as many 'tracks' or strands as possible, starting with two, then four, doubling ultimately to 32. It is common practice for the same initial set of strands on a master to be re-mixed. To the end user the sound is a complete, seamless product but it can always be broken down and differently re-constituted.
It is much more difficult to manipulate the data on an ordinary photograph. It can be proportionately enlarged or the colours can be uniformly altered but it is difficult to manipulate the content asymetrically.
Digital information should be designed much more like a CD than a photograph. This can only be achieved if the information is stratified and labelled. Instead of mixing information like two different colours of paint in one pot (so that the two colours cannot be subsequently separated), each piece of information should be thought of as a layer or stratum. So, if you think of a blue on grey line drawing of York Minster with its name in red, even if you have a good set of customisation tools for contrast and colour your scope is limited because you have three colours and you can only exercise the contrast over all elements simultaneously. If you label the line drawing "Image" and the name "title" and place the latter 'on top' of the former then the two pieces of information can be separated by the user who can then, if desired:
- View the drawing without the title
- View the title without the drawing
- Alter the contrast of each separately
- Adjust the content of each separately, e.g. alter print font, reduce print size without reducing drawing scale.
Obviously, the more complex an image the more important it is to produce and render it in layers.
These procedures present genuine problems relating to the integrity of intellectual property. Processes need to be incorporated which allow the creation of temporary files for manipulation but these must be easily distinguishable from the source file and they must have a limited 'life' and be usable on one device only during their 'lifetime'.
3.2 The Power of Nine
Although a great many interoperability features can be automated (se 3.9 below) through XML and style sheets, one area which requires a high degree of human intervention is taxonomy. In The Power of Nine (Carey & Stringer, Library and Information Commission, 2000. Report no.74 (now Re:Source), 2000) there are two basic arguments advanced in favour of taxonomy by nine: first, this observes the optimal "7 + or - 2" selection rule; secondly, it accords with the numeric keypads for consumer electronics remote controllers, telephones and standard PC keyboards.
humanITy has developed some basic rules for applying the general theory and has assembled working notes resulting from its experience of adapting a community information site (East Sussex Community Information Service – ESCIS, www.escis.org.uk) which initially posted 18 top level information categories. Taxonomy is not in itself a new or even a primitive discipline but it is not a skill usually associated with Web designers who are more concerned with aesthetics.
In using any information system there has to be an optimal balance between links and 'clicks'. Although tests have not yet been undertaken comparing information systems using the Power of Nine and those using a smaller or larger number of links per display, it is likely that in many circumstances the Power of Nine may be slower to use than systems with more links per page. The speed of access needs to be balanced with he mode of access; e.g. a system where more than nine links appear on one array will not be numeric keypad accessible.
The Power of Nine is not indefinitely scaleable; the optimal number of 'clicks' is probably only three or four.
In an environment where interoperability is important spatial navigation, based on nine or otherwise, needs to be only one option alongside alphanumeric, chronological, key-word &c. Image maps, however, should be avoided as they are inaccessible to numeric keypad systems and to access devices which work with text. It is also vital to develop interrogative systems to assist searches, particularly where these are being conducted on a telephone without a screen. "What do you want?" tied to a specialist voice-in micro vocabulary may be a lot more convenient for consumer and producer than a series of menus. If, for example, you are phoning a health line, there is only a fairly restricted number of medical terms with which the caller will be familiar. This relates to 3.3 and 3.4 below.
3.3 Discrete and Multi-media Access
One of the strange disjunctions between televisions and PCs is the unwritten rule of access whereby television is assumed to be multimedia by nature (pictures, sound, text presented simultaneously) whereas PC access is assumed to be vision only (pictures and text). Recently, for example, the DfES commissioned research on the comparative efficacy of analogue and digital devices for the improvement of literacy in adults. The digital aspect of the testing assumed that adults would simply have access to screens displaying symbolic language (text) but they would not have access to complementary speech output; why not? The answer is an obsession with process over output. This may well be a legitimate concern of educators but should not be a self-imposed restriction by information and entertainment providers.
As a basic principle, all content should be rendered in as many media as possible but each medium should be self-standing. The first part of this principle is obvious, the second requires a little explanation. For the purposes of accessibility people with severe hearing difficulties require sub-titling and people with severe sight difficulties may require additional audio to that supplied by standard products; this involves writing a script adapted from the original which provides sub titles which simplify dialogue (see 2.4 below) and also requires 'stretched' stage directions as additional description during parts of a video presentation where there is no dialogue. In cases of severe impairment, deaf people require an information stream which is totally visual and is not dependent in any way on audio clues and the converse is true of blind people who need an audio stream which does not rely upon any visual element.
These specialist applications now have to be extended because of access to the web by PDAs and mobile phones. PDAs and some phones will have small screens which require audio support to graphic displays but automobile output systems, digital radios and some telephones will require information which does not rely at all upon visual output.
As with the remarks on taxonomy, creating information for interoperability of this sort relies upon a high degree of abstract thinking and language skills. A designer needs to know, for example, the difference between what something is and what it does, otherwise the audio description of visual elements might be completely misleading.
3.4 Language Engineering and Simplification Engines/Tools
The concept of engineering between languages is well established and it is of major relevance to community information sites, particularly those which serve ethnic/linguistic minorities. Crucially, the language options should not be presented in English Roman script. If you only know Urdu you may not know the word "Urdu".
In the context of a functional illiteracy rate of over 20% for standard analogue tests, with the likelihood that the rate is no less in digital information access, language engineering within languages is particularly important, in tandem with speech out assistance. The two major elements are lexicographic tranching and précis.
Lexicographic tranching is familiar in two environments at different ends of the language spectrum. At one end, tabloid newspapers run their texts through tools in order to establish a given 'reading age' for the text. Tools analyse the lexicography and flag words outside/'above' the reading age limit. A Thesaurus-type function can then be used to suggest substitutes. At the other end of the spectrum there is lexicographic analysis for literature studies which, for example, counts the use of given words or variants in a corpus of text such as the plays of Shakespeare (vid: Hubert, J.J.: Linguistic indicators. Social Indicators Research 8 (1980) pp. 223-255. Holland and Boston, USA). Frequency analysis of this sort is a good supportive indicator to the 'reading age' tests.
It is easy to over simplify these approaches. In presenting public information it is likely that there needs to be a distinction between general vocabulary and specialist micro vocabularies. To take two instances: many tabloid reading males have a larger soccer micro vocabulary than the whole of their general vocabulary; a popularised version of a Government policy on tenancy may require a micro vocabulary of terms connected with housing and associated legal processes.
To shorten is not necessarily to simplify; an executive summary of a technical document may be more difficult to read than the full-length document because the arguments and the lexicography are telescoped (e.g. through acronyms).
As far as we are aware, these tools are not available on any Government information website. What we would propose is as follows:
Options on a two-dimensional matrix with a percentage of text on the x axis and lexicographic level on the y axis. Instead of describing the y-axis as "Reading age" one might use the three terms: "Simple, General" and "Professional".
Précis engines are developing rapidly. They are obviously useful for the x-axis function. In respect of x and y functions consideration must be given to the role of these tools vis à vis the producer and consumer. The producer should think of the market and use these tools in order, perhaps, to give the user two or three different versions of a document with the one preferred by the consumer as the lead option, with a menu indicating other options. This is particularly important where the consumer does not have lexicographic or précis tools. These will become increasingly available on PCs but are not likely to be available in the near future on PDAs and telephones. In these cases 'server side' operations might be invoked but all of this is very complex.
3.5 Integration of Procedure and Processing Tools into Systems
Here are two simple illustrations:
- Inland Revenue. I wish to complete my Income Tax Return online. As I struggle with the questions I need to refer to the notes that are not cross-referenced in the form itself and actually appear separately as in a paper copy.
- Passport Application. In applying for a Passport because I have had my current Passport stolen, I am asked for the number of my current Passport.
Tools for data matching and customer prompting should be utilised by the producer. A good example of this is the number of products, which match postcodes with names, addresses, and telephone numbers. A user who gives a surname, given names and a post code should be spared a great deal of bother, not having to write an address or a history which is already available to the producer. In the Passport example, the Home Office already knows the number of the current Passport because it issued it. Why should it ask the consumer? If we are to run the 'Big Brother' risk, it is important not to annoy the consumer by wasting time. The tax example is a good illustration of why it is not enough to take a paper document and reproduce it in exactly the same form on a screen. A good example of integration is the inclusion of calculator tools in e-shopping sites. An example of bad practice is the separation of buying and paying on e-shopping sites.
Some systems offer a choice of "Help" levels but many do not; and even more have no "Help" facility. Ideally, "Help" should be heuristic, adapting to the 'known' needs of the individual user. Whatever the sophistication of the system, the producer should establish an optimal level of "Help" for a system and integrate tools and prompts into it.
3.6 All-to-All Processing Systems and Collaborative Process and Procedure Tools
Most information systems work like valves; they assume that transactions are one-way, usually from producer to consumer.
When people use the term "Interactive" in the context of digital media they usually mean the ability of the consumer to choose between offers made by the producer, e.g. to choose a soccer game or to choose camera angles within the soccer game. They do not usually mean the ability of the consumer to address the producer. In other words, we have a “one to all” message system. At the very least a good information system should have an “all to one” facility.
However, the great virtue of the Internet is that it facilitates “all to all” message systems. The most popular form of these is the e-mail discussion list. This can work really well under the following two conditions: first, that it follows (does not precede) human interaction; secondly, that the people involved have a clear idea of their shared agenda and the limits of the list's functions and competences. To the extent to which these two conditions do not apply, e-mail lists in particular and “all-all” message systems in general are proportionately difficult to operate effectively. Unless there is a very clear shared agenda, e-list discipline easily breaks down and requires a moderator. This is a particular aspect of the difficulty of hybrid systems.
In community information systems it is important to be sensitive to the potential and problems of “all-all” systems. They generally require a good deal of input concerning process (how the system performs a function) and procedure (how individuals relate to the process). In other words, people need to know how systems work and they need to know how they relate to the system and the other people using it. This last point is very important if people are trying to work together. This is an extremely difficult area that is much more than 'netiquette'.
There is a whole area of research, which needs to be conducted into hybrids between digital information and human information facilitation. At the one extreme there is the human being with a body of knowledge, at the other extreme there is the totally automated, digitised system. Between them there are phenomena such as the call centre (routing system > human being > computer database) and the enquiry line (routing system > data location system > human assistance option). The use of hybrids for on-screen information is much less well developed but the example of the facilitated list is well known (data entry > human editor > distribution).
One final point needs to be made here. One, sometimes explicit, sometimes not, assumption of the public sector is that automated, digitised data systems will cut the cost of citizen/state transactions. This is doubtful: if the transaction experience is poor it will not reduce human transactions; if the experience is good is will increase the number of both 'virtual' and human transactions.
3.7 Multiple Label and Multiple Attribute Allocation to Data
The idea of multiple attributes for data is at once intuitively easy to grasp but surprisingly difficult to implement. We are all aware that a piece of data which we are handling might be classified in a variety of ways but, apart from the simple cross-reference, we are content to give the data a single attribute. This naturally arises from the difficulty of multiple attributions in analogue formats. It accounts, for instance, for the development of the Dewey System which determines where a book should be shelved when it might be accorded a variety of attributes;
there is only one copy of the book, so that is where it has to be. This is obviously not the case with hypertext. Indeed, the capacity to allocate multiple attributes may lead to problems with over 'stuffed' search reports where core data and 'incidental' data are listed together. Nonetheless, in spite of this potential problem, in community information systems in particular multiple attributes are very important. You might, for example, want to classify a "Church" under the primary classification of "Belief" but many users might want to reach the facility for the purposes of philanthropic support or because the church in question might have a school. "Churches" would therefore need to appear as a sub classification in a list of community organisations and in a list of schools.
The idea of multiple labelling is a 'mirror' of the above. IN this case a concept is given a different label depending on its context; so, for example, in a part of a community information site dedicated to health care the sub-classification "Jabs" may appear but in a more medically oriented database the term "Vaccination" might be used. This might lead to the same piece of information.
So, in summary, there should be a variety of routes to the same information and there should, where appropriate, be a variety of labels for information.
3.8 Uniform Customisation
Because of the way in which information is designed, customisation is rarely uniform.
The most obvious form of customisation is the ability to alter foreground and background colour, contrast and the size and font of print within arrays of data. This, however, is not yet common practice in web design even if it is a familiar feature of word processors. Authorial aesthetics play a large part in rendering data inaccessible and not customised so, to preserve appearance authors often omit the facility to enlarge text and change its font but even more often page authors define the pixel dimensions for images so that these cannot be 'blown up'.
The most serious problem of all in customisation is failure to facilitate it in such metadata as home pages and menus. It does not matter how accessible the data is if you can't locate it because of an inability with metadata.
3.9 XML with Style Sheets
Many fundamental accessibility problems are caused by the employment by designers of a huge variety of tools and attributions, some highly proprietary, other open source or industry standard. To take just one example of an 'open standard'; HTML has been in widespread use for a number of years but its development has far outstripped the standard setting of the World Wide Web Consortium (www.w3.org) which means that it has been bent and twisted by web authors to produce special effects which break the universality.
The kind of problems raised in this Report can only be resolved with a degree of standardisation. Such discipline also improves interoperability.
There are a large variety of technical problems presented by the concept of interoperability, particularly if information is dynamic, requiring updating for a variety of formats - PC/Web, Television/Web, mobile phone - in which case the solution must be to use XML with a set of output-specific style sheets.
There are some technical notes at Appendix 1.
Appendix 1.
Technical Notes on Information for Interoperability
NB this is not a glossary as such, it is a set of notes about the way items are used.
- AvantGo <https://avantgo.com/frontdoor/index.html>
- Off line browser for small devices.
- Browsers
- Initially Netscape, Lynx and Mosaic via PC (standard screen and resolution). Phones, PDAs, &c, new browser technologies, no standard screen size/resolution
- Connection Speed
- PDA, Web/TV, mobile phones slow; WAP also limits page storage size. PDAs, palmtops often download via other devices which slows them down
- Devices/Appliances/Receivers
- Brief list of information devices to bear in mind when creating information:
- Apple Mackintosh
- Browsers - desktop, pocket and text-based; Palm, Symbian/epoc; off-line
- Games consoles
- I-Mode phones
- Screen readers (for telephone access or visually impaired people)
- WAP phones
- Webpads (pen operated)
- Web TV
- Windows PC
- Graphics
- Keep to minimum or dispense with for information on small and slow devices.
- Guidelines for Interoperability, Mobile HTML
-
- HTML 3.2
- Pages 10kb or less
- Cookies
- Tag for handheld or Palm friendly
- Images maximum 153 pixels wide for images/tables.
- HTML (Hypertext mark-up language)<www.w3.org>
- HTML has been 'bent' to change it from a descriptor language, separating style from content, to a prescriptive layout tool. The more prescriptive the less interoperable.
- Image Maps
- Visual navigation aids no use with text only browsers or small screen devices.
- Javascript:<http://www.jsworld.com>
- By no means universally accessible and works unreliably.
- Palm.Net
- Provides Web Clipping service for mobile phones via a proxy server, based on html not wml.
- Plug-ins
- Work much better on PCs than small devices. Plug-ins for web editors such as Dreamweaver and GoLive (or Go-Live) supplied by Nokia, <http://forum.nokia.com>.
- Pointers
- Many devices don't have them.
- Vignette
- Useful reference site for leading edge data serving, used by the Guardian; uses templates matched to receiver device.
- WAI (Web Accessibility Initiative)
- Best known for its Web Content Accessibility Guidelines (WCAG) but more significant in the long run for guidelines on authoring tools.
- WAP (Wireless Access Protocol)
- Usually slower on download than advertised.
- WAP Activ Server
- Provides tools for mobile phone users, including adapting WML to match mobile phones.
- WCA (Web Clipping Application):
- Free download.
- Web Editors
- Applications software for writing in HTML, e.g. Macromedia Dreamweaver, Adobe Go-Live.
- WML (Wireless Mark-up Language)
- Code for describing wireless data.
- XML (Extensible Mark-up Language)
- The holy grail, for now. Separates style from content much better than HTML by using style sheets. The single source file data is updated; re-design is achieved through changing the style sheets for each device.