Bill Cope and Mary Kalantzis

The discursive practice of markup

Untagged digital text is a linear data-type, a string of sequentially ordered characters. Such a string is regarded as unmarked text. Markup adds embedded codes or tags to text, turning something which is, from the point of view of computability, a flat and unstructured data-type, into something which is potentially computable (Buzzetti and McGann 2006).

Tagging or markup does two things. First, it does semantic work. It notates a text (or image, or sound) according to a knowledge representation of its purported external referrents. It connects a linear text with a formal conceptual scheme. Second, it does structural work. It notates a text with one particular semantics, the semantics of text itself, referring specifically to the text’s internal structures.

In the first, semantic agenda, a tag schema constitutes a controlled vocabulary describing a particular field in a formally defined and conceptually rigorous way. The semantics of each tag is defined with as little ambiguity as possible in relation to the other tags in a tag schema. Insofar as the tags relate to each other—they are indeed a language—they can be represented by means of a tag schema making structural connections (a < Person > is named by < GivenNames > and < Surname >) and counter distinctions against each other (the < City > Sydney as distinct from the < Surname > of the late eighteenth-century British Colonial Secretary after which the city was named). Schemas define tags paradigmatically.

To take Kant’s example of the willow and the linden tree, and express it the way a tagging schema might, we could mark up these words semantically as < tree > willow </tree > and < tree > linden tree </tree >. The tagging may have a presentational effect if these terms need highlighting, if they appear as keywords in a scientific text, for instance; or it may assist in search. This markup tells us some things about the structure of reality, and with its assistance we would be able to infer that a < tree > beech </tree > falls into the same category of represented meaning. Our controlled vocabulary comes from somewhere in the field of biology. In that field, a < tree > is but one instance of a < plant >. We could represent these structural connections visually by means of a taxonomy. However, < tree > is not an unmediated element of being; rather, it is a semantic category. How do we create this tag-category? How do we come to name the world in this way?

Eco (1999) provides Kant’s answer:

What follows is a process Kant calls ‘the legislative activity of the intellect’. From the intuition of trees, the intellect creates the concept of tree. ‘[T]o form concepts from representations it is… necessary to be able to compare, reflect, and abstract; these three logical operations of the intellect, in fact, are the essential and universal conditions for the production of any concept in general’ (quoted in Eco 1999, pp. 74–75).

Trees exist in the world. This is unexceptionable. We know they exist because we see them, we name them, we talk about them. We do not talk about trees because they are mere figment of conceptual projection, the result of a capricious act of naming. There is no doubt that there is something happening, ontologically speaking. However, we appropriate trees to thought, meaning, representation and communication through mental processes which take the raw material of sensations and from these construct abstractions in the form of concepts and systems of concepts or schemas. These concepts do not transparently represent the world; they represent how we figure the world to be.

And how do we do this figuring? When we use the concept ‘tree’ to indicate what is common to willows, linden trees and beeches, it is because our attention has been fixed on specific, salient aspects of apprehended reality—what is similar (though not the same) between the two trees, and what is different from other contiguous realities, such as the soil and the sky. But equally, we could have fixed our attention on another quality, such as the quality of shade, in which respect a tree and a built shelter share similar qualities.

Tags and tag schemas build an account of meaning through mental processes of abstraction. This is by no means an ordinary, natural or universal use of words. Vygotsky and Luria make a critical distinction between complex thinking and conceptual thinking. Complex thinking collocates things that might typically be expected to be found together: a tree, a swing, grass, flower beds, a child playing and another tree—for their circumstantial contiguity, the young child learns to call these a playground. From the point of view of consciousness and language, the world hangs together through syncretic processes of agglomeration. A playground is so named because it is this particular combination of things. The young child associates the word ‘playground’ with a concrete reference point. Conceptual thinking also uses a word, and it is often the same word as complex thinking. However, its underlying cognitive processes are different. Playground is defined functionally, and the word is used ‘as a means of actively centring attention, of abstracting certain traits, and symbolising them by the sign’ (Vygotsky 1962; also referred to in Cope and Kalantzis 1993 and Luria 1981).

Then, beyond the level of the word-concept, a syntax of abstraction is developed in which concept relates to concept. This is the basis of theoretical thinking, and the mental construction of accounts of a reality underlying what is immediately apprehended, and not even immediately visible (Cope and Kalantzis 1993). The way we construct the world mentally is not just a product of individual minds; it is mediated by the acquired structures of language with all its conceptual and theoretical baggage—the stuff of socialised worldviews and learned cultures.

Conceptual thinking represents a kind of ‘reflective consciousness’ or metaconsciousness. Markup tags are concepts in this sense and tag schemas are theories that capture the underlying or essential character of a field. When applied to the particularities of a specific piece of content, they work as a kind of abstracting metacommentary, relating the specifics of a piece of content to the generalised nature of the field. Academic, disciplinary work requires a kind of socio-semantic activity at a considerable remove from commonsense associative thinking.

Markup tags do not reflect reality in an unmediated way, as might be taken to be the case in a certain sense of the word ‘ontology’. Nor do they represent it comprehensively. Rather, they highlight focal points of attention relevant to a particular expressive domain or social language. In this sense, they represent worldviews. They are cultural artefacts. A tag does not exhaustively define the meaning function of the particular piece of content it marks up. Rather, it focuses on a domain-specific aspect of that content, as relevant to the representational or communicative purposes of a particular social language. In this sense ‘schema’ is a more useful concept than ‘ontology’, which tends to imply that our representations of reality embody unmediated truths of externalised being.

Notwithstanding these reservations, there is a pervasive underlying reality, an ontological grounding, which means that schemas will not work if they are mere figments of the imagination. Eco characterises the relationship between conceptualisation and the reality to which it refers as a kind of tension. On the one hand

But this does not mean that anything goes. ‘We learn by experience that nature seems to manifest stable tendencies… [S]omething resistant has driven us to invent general terms (whose extension we can always review and correct)’ (Eco 1999). The world can never be simply a figment of our concept-driven imaginations. ‘Even granting that the schema is a construct, we can never assume that the segmentation of which it is the effect is completely arbitrary, because… it tries to make sense of something that is there, of forces that act externally on our sensor apparatus by exhibiting, at the least, some resistances’ (Eco 1999). Or as Latour says of the work of scientists, the semantic challenge is to balance facticity with social constructivism of disciplinary schemata (Latour 2004).

Structural markup

Of the varieties of textual semantic markup, one is peculiarly self-referential, the markup of textual structure, or schemas which represent the architectonics of text. A number of digital tagging schemas have emerged, which provide a functional account of the processes of containing, describing, managing and transacting text. They give a functional account of the world of textual content. Each tagging schema has its own functional purpose. A number of these tagging schemas have been created for the purpose of describing the structure of text, and to facilitate its rendering to alternative formats. These schemas are mostly derivatives of SGML, HTML and XHTML, and are designed primarily for rendering transformations through web browsers. Created originally for technical documentation, DocBook structures book text for digital and print renderings. The Text Encoding Initiative is ‘an international and interdisciplinary standard that helps libraries, museums, publishers, and individual scholars represent all kinds of literary and linguistic texts for online research and teaching’ (http://www.tei-c.org/).

Although the primary purpose of each schema may be a particular form of rendering, this belies the rigorous separation of semantics and structure from presentation. Alternative stylesheet transformations could be applied to render the marked up text in a variety of ways on a variety of rendering devices.

These tagging schemas do almost everything conceivable in the world of the written word. They can describe text comprehensively, and they support the manufacture of variable renderings of text on the fly by means of stylesheet transformations. The typesetting and content capture schemas provide a systematic account of structure in written text, and through stylesheet transformations they can render text to paper, to electronic screens of all sizes and formats, or to synthesised audio.

Underlying this is a fundamental shift in the processes of text work, described in Chapter 4, ‘What does the digital do to knowledge making?’. A change of emphasis occurs in the business of signing—broadly conceived as the design of meaning—from configuring meaning form (the specifics of the audible forms of speaking and the visual form of written text) to ‘marking up’ for meaning function in such a way that alternative meaning forms, such as variable visual (written) and audio forms of language, can be rendered by means of automated processes from a common digital source.

In any digital markup framework that separates the structure from presentation, the elementary unit of meaning function is marked by the tag, specifying the meaning function for the most basic ‘chunk’ of represented content. Tags, in other words, describe the meaning function of a unit of content. For instance, a word or phrase may be tagged as < Emphasis >, < KeywordTerm > or < OtherLanguageTerm >. These tags describe the peculiar meaning function of a piece of content. In this sense, a system of tags works like a functional grammar; it marks up key features of the information architecture of a text. Tags delineate critical aspects of meaning function, and they do this explicitly by means of a relatively consistent and semantically unambiguous metalanguage. This metalanguage acts as a kind of running commentary on meaning functions which are otherwise embedded, implicit or to be inferred from context.

Meaning form follows mechanically from the delineation of meaning function, and this occurs in a separate stylesheet transformation space. Depending on the stylesheet, for instance, each of the three functional tags < Emphasis >, < KeywordTerm > and < OtherLanguageTerm > may be rendered to screen or print either as boldface or italics, or as an audible intonation in the case of rendering as synthesised voice.

Given the pervasiveness of structural markup, one might expect that an era of rapid and flexible transmission of content would quickly dawn. But this has not occurred, or at least not yet, and for two reasons. The first is the fact that, although almost all content created over the past quarter of a century has been digitised, the formats are varied and incompatible. Digital content is everywhere, but most of it has been created, and continues to be created, using typographically oriented markup frameworks. These frameworks are embedded in software packages that provide tools for working with text which mimic the various trades of the Gutenberg universe: an author may use Word; a desktop publisher or latter-day typesetter may use inDesign; and a printer will use a PDF file as if it were a virtual forme or plate. The result is sticky file flow and intrinsic difficulties in version control and digital repository maintenance. How and where is a small correction made to a book that has already been published? Everything about this relatively simple problem, as it transpires, remains complex, slow and expensive. However, in a fully comprehensive, integrated file flow, things that are slow and expensive today should become easier and cheaper—a small change by an author to the source text could be approved by a publisher so that the very next copy of that work could include that change.

To return to the foundational question of the changed means of production of meaning in semantic and structural text-work environment, we want to extend the distinction of ‘meaning form’ from ‘meaning function’. Signs are the elementary components of meaning. And ‘signs’, say Kress and Leeuwen, are ‘motivated conjunctions of signifiers (forms) and signifieds (meanings)’ (Kress and van Leeuwen 1996). Rephrasing, we would call motivated meanings, the products of the impulse to represent the world and communicate those representations, ‘meaning functions’. The business of signing, motivated as it is by representation (meaning interpreted oneself) and communication (meaning interpreted by others), entails an amalgam of function (a reason to mean) and form (the use of representational resources which might adequately convey that meaning).

The meaning function may be a flower in a garden on which we have fixed our focus for a moment through our faculties of perception and imagination. For that moment, this particular flower captures our attention and its features stand out from its surroundings. The meaning function is our motivation to represent this meaning and to communicate about it. How we represent this meaning function is a matter of meaning form. The meaning form we choose might be iconic—we could draw a sketch of the flower, and in this case, the act of signing (form meets function) is realised through a process of visual resemblance. Meaning form—the drawing of the flower—looks like meaning function, or what we mean to represent: the flower. Or the relation between meaning form and function may be, as is the case for language, arbitrary. The word ‘flower’, a symbolic form, has no intrinsic connection with the meaning function it represents. In writing or in speech the word ‘flower’ conventionally represents this particular meaning function in English. We can represent the object to ourselves using this word in a way which fits with a whole cultural domain of experience (encounters with other flowers in our life and our lifetime’s experience of speaking about and hearing about flowers). On the basis of this conventional understanding of meaning function, we can communicate our experience of this flower or any aspect of its flower-ness to other English speakers.

This, in essence, is the stuff of signing, the focal interest of the discipline of semiotics. It is an ordinary, everyday business, and the fundamental ends do not change when employing new technological means. It is the stuff of our human natures. The way we mean is one of the distinctive things that makes us human.

One of the key features of the digital revolution is the change in the mechanics of conjoining meaning functions with meaning forms in structural and semantic markup. We are referring here to a series of interconnected changes in the means of production of signs. Our perspective is that of a functional linguistics for digital text. Traditional grammatical accounts of language trace the limitlessly complex structures and patterns of language in the form of its immediately manifest signs. Only after the structure of forms has been established is the question posed, ‘what do these forms mean?’. In contrast, functional linguistics turns the question of meaning around the other way: ‘how are meanings expressed?’. Language is conceived as a system of meanings; its role is to realise or express these meanings. It is not an end in itself; it is a means to an end (Halliday 1994). Meaning function underlies meaning form. An account of meaning form must be based on a functional interpretation of the structures of meaning. Meaning form of a linguistic variety comprises words and their syntactical arrangement, as well as the expressive or presentational processes of phonology (sounding out words or speaking) and graphology (writing). Meaning form needs to be accounted for in terms of meaning function.

Structural and semantic markup adds a second layer of meaning to the process of representation in the form of a kind of meta-semantic gloss. This is of particular value in the deployment of specialised disciplinary discourses. Such discourses rely on a high level of semantic specificity. The more immersed you are in that particular discourse—the more critical it is to your livelihood or identity in the world, for instance—the more important these subtle distinctions of meaning are likely to be. Communities of practice identify themselves by the rigorous singularity of purpose and intent within their particular domain of practice, and this is reflected in the relative lack of terminological ambiguity within the discourse of disciplinary practice of that domain. In these circumstances semantic differences between two social languages in substantially overlapping domains is likely to be absolutely critical. This is why careful schema mapping and alignment is such an important task in the era of semantic and structural markup.

Metamarkup: developing markup frameworks

We now want to propose in general terms an alternative framework to the formalised semantic web, a framework that we call an instance of ‘semantic publishing’. The computability of the web today is little better than the statistical frequency analyses of character clusters that drive search algorithms, or the flat and featureless world of folksonomies and conceptual popularity rankings in tag clouds. Semantic publishing is a counterpoint to these simplistic modes of textual computability. We also want to advocate a role for conceptualisation and theorisation in textual computability, against an empiricism which assumes that the right algorithms are all we need to negotiate the ‘data deluge’—at which point, it is naively thought, all we need to do is calculate and the world will speak for itself (Anderson 2009). In practical terms, we have been working through and experimenting with this framework in the nascent CGMeaning online schema making and schema matching environment.

Developing an interlanguage mechanism

By filtering schemata through the ‘interlanguage’ mechanism, a system is created that allows conversation and information interchange between disjoint schemas. In this way, it is possible to create functionalities for data framed within the paradigm of one schema which extend well beyond those originally conceived by that schema. This may facilitate interoperability between schemas, allowing data originally designed for use in one schema for a particular set of purposes to be used in another schema for a different set of purposes.

The interlanguage mechanism means that metadata newly created through its apparatus to be interpolated into any number of metadata schemas. It also provides a method by means of which data harvested in one metadata schema can be imported into another. From a functional point of view, some of this process can be fully automated, and some the subject of automated queries requiring a human-user response.

The interlanguage mechanism, in sum, is designed to function in two ways:

The filter apparatus is driven by a set of semantic and syntactical rules as outlined below, and throws up queries whenever an automated translation of data is not possible in terms of those semantic rules.

The interlanguage apparatus is designed to be able to read tags, and thus interpret the data which has been marked up by these tags, according to two overarching mechanisms, and a number of submechanisms. The two overarching mechanisms are the superordination mechanism and the composition mechanism—drawing in part here on some distinctions made in systemic-functional linguistics (Martin 1992).

The superordination mechanism constructs tag-to-tag ‘is a…’ relationships. Within the superordination mechanism, there are the submechanisms of hyponymy (‘includes in its class…’), hyperonymy (‘is a class of…’), co-hyperonoymy (‘is the same as…’), antinomy (‘is the converse of…’) and series (‘is related by gradable opposition to…’).

The composition mechanism constructs tag-to-tag ‘has a.’ relationships. Within the composition mechanism, there are the submechanisms of meronymy (‘is a part of…’), co-meronymy (‘is integrally related to but exclusive of…’), consistency (‘is made of…’), collectivity (‘consists of…’).

These mechanisms can be fully automated in the case of new data formation within any schema, in which case, deprecation of some aspects of an interoperable schema may be required as a matter of course at the point of data entry. In the case of legacy data generated in schemas without anticipation of, or application of, the interlanguage mechanism, data can be imported in a partially automated way. In this case, tag-by-tag or field-by-field queries are automatically generated according to the filter mechanisms of:

The interlanguage mechanism (Figure 13.1) is located in CGMeaning, a schema-building and alignment tool. This software defines and determines:

Following are some examples of how this mechanism may function. In one scenario, new data might be constructed according to a source schema which has already become ‘aware’ by means of previous applications of the interlanguage mechanism as a consequence of the application of the mechanism. In this case, the mechanism commences with the automatic interpellation of data, as the work of reading and querying the source schema has already been performed. In these circumstances, the source schema in which the new data is constructed becomes a mere facade for the interlanguage, taking the form of a user interface behind which the processes of subordination and composition occur.

In another scenario, a quantum of legacy source data is provided, marked up according to the schematic structure of a particular source schema. The interlanguage mechanism then reads the structure and semantics immanent in the data, interpreting this from schema and the way the schema is realised in that particular instance. It applies four filters: a delicacy filter, a synonymy filter, a contiguity filter and a subset filter. The apparatus is able to read into the schema and its particular instantiation an inherent taxonomic or schematic structure. Some of this is automated, as the relationships of tags is unambiguous based on the readable structure of the schema and evidence drawn from its instantiation in a concrete piece of data. The mechanism is also capable of ‘knowing’ the points at which it is possible there might be ambiguity, and in this case throws up a structured query to the user. Each human response to a structured query becomes part of the memory of the mechanism, with implications drawn from the user response and retained for later moments when interoperability is required by this or another user. On this basis, the mechanism interpellates the source data into the interlanguage format, while at the same time automatically ‘growing’ the interlanguage itself based on knowledge acquired in the reading of the source data and source schema.

Having migrated into the interlanguage format, the data is then reworked into the format of the destination schema. It is rebuilt and validated according to the mechanisms of superordination (hyponymy, hyperonymy, co-hyperonomy, antonymy and series) and composition (meronymy, co-meronymy, consistency, collectivity). A part of this process is automated, according to the inherent structures readable into the destination schema, or previous human readings that have become part of the accumulated memory of the interlanguage mechanism. Where the automation of the rebuilding process cannot be undertaken by the apparatus with assurance of validity (when a relation is not inherent to the destination schema, nor can it be inferred from accumulated memory in which this ambiguity was queried previously), a structured query is once again put to the user, whose response in turn becomes a part of the memory of the apparatus, for future use. On this basis, the data in question is interpolated into its destination format. From this point, it can be used in its destination context or schema environment, notwithstanding the fact that the data had not been originally formatted for use in that environment.

Key operational features of this mechanism include:

In the most challenging of cases—in which the raw digital material is created in a legacy schema, and in which that schema is not already known to the interlanguage from previous interactions—the mechanism:

These inferences then become visible to subsequent users, and capable of amendment by users, through the CGMeaning interface, which:

To give another example, the source schema is already known to the interlanguage, by virtue of automated validations based not only on the inherent structure of the schema, but also many validations against a range of data instantiations of that schema, and numerous user clarifications of queries. In this case, by entering data in an interface that ‘knowingly’ relates to an interlanguage which has been created using the mechanisms provided here, there is no need for the filter mechanisms nor the interpolation processes that are necessary in the case of legacy data and unknown source schemas; rather, data is entered directly into the interlanguage format, albeit through the user interface ‘facade’ of the source schema. The apparatus then interpolates the data onto the designated destination schema.

Schema alignment for semantic publishing: the example of Common Ground Markup Language

Common Ground Markup Language (CGML) is a schema for marking up and storing text as structured data, created in the tag definition and schema alignment environment, CGMeaning. The storage medium can be XML files, or it can be a database in which fields are named by tags, and from which exports produce XML files marked up for structure and semantics, ready for rendering through available stylesheet transformations. The result is text that is more easily located by virtue of the clarity and detail of metadata markup, and capable of a range of alternative renderings. CGML structures and stores data on the basis of a functional account of text, not just as an object but as a process of collaborative construction. The focal point of CGML is a functional grammar of text, as well as a kind of grammar (in the metaphorical sense of generalised reflection) of the social context of text work. However, with CGML, ‘functional’ takes on a peculiarly active meaning. The markup manufactures the text in the moment of rendering, through the medium of stylesheet transformation in one or several rendering processes or media spaces.

In CGML, as is the case for any digital markup framework that separates structure and semantics from presentation, the elementary unit of meaning function is marked by the tag. The tag specifies the meaning function for the most basic ‘chunk’ of represented content. Tags, in other words, describe the meaning function of a unit of content. For instance, a word or phrase may be tagged as < Emphasis >, < KeywordTerm > or < OtherLanguageTerm >. These describe the peculiar meaning function of a piece of content. In this sense, a system of tags works like a partial functional grammar: they mark up key features of the information architecture of a text. Tags delineate critical aspects of meaning function, and they do this explicitly by means of a relatively consistent and semantically unambiguous metalanguage. This metalanguage acts as a kind of running commentary on meaning functions, which are otherwise embedded, implicit or to be inferred from context.

Meaning form follows mechanically from the delineation of meaning function, and this occurs in a separate stylesheet transformation space. Depending on the stylesheet, each of the three functional tags < Emphasis >, < KeywordTerm > and < OtherLanguageTerm > may be rendered to screen or print either as boldface or italics, or as a particular intonation in the case of rendering as synthesised voice. Stylesheets, incidentally, are the exception to the XML rule strictly to avoid matters of presentation; meaning form is their exclusive interest.

CGML, in other words, is a functional schema for authorship and publishing. CGML attempts to align the schemas we will describe shortly, incorporating their varied functions. CGML is an interlanguage. Its concepts constitute a paradigm for representational work, drawing on a historically familiar semantics, but adapting this to the possibilities of the internet. Its key devices are thesaurus (mapping against functional schemas) and dictionary (specifying a common ground semantics). These are the semantic components for narrative structures of text creation, or the retrospective stories that can be told of the way in which, for instance, authors, publishers, referees, reviewers, editors and the like construct and validate text. The purpose of this work is both highly pragmatic (such as a description of an attempt to create a kind of functional grammar of the book) and highly theoretical (a theory of meaning function capable of assisting in the partially automated construction and publication of variable meaning forms).

In the era of digital media, the social language of textuality is expressed in a number of schemas. It is increasingly the case that these schemas perform a wide ranging, fundamental and integrated set of functions. They contain the content—the electronic files that provide structural and semantic shape for the data which will be rendered as a book. They describe the content—for the purposes of data transfer, warehousing and retrieval. They manage the content—providing a place where job process instructions and production data are stored. And they transact the content.

A number of digital tagging schemas have emerged which provide a functional account of these processes of containing, describing, managing and transacting text. More broadly, they provide a functional account of the world of textual content in general. Each tagging schema has its own functional purpose, or ‘funnel of commitment’, to use Scollon’s terminology. We will briefly describe a few of these below, categorising them into domains of professional and craft interest: typesetting and content capture, electronic rendering, print rendering, resource discovery, cataloguing, educational resource creation, e-commerce and digital rights management. The ones we describe are those we have mapped into CGML.

What tagging schemas do

The tagging schemas we have mentioned here do almost everything conceivable in the world of the written word. They can describe that world comprehensively, and to a significant degree they can support its manufacture. The typesetting and content capture schemas provide a systematic account of structure in written text, and through stylesheet transformations they can literally print text to paper, or render it electronically to screen or manufacture synthesised audio. Digital resource discovery and electronic library cataloguing schemas provide a comprehensive account of the form and content of non-digital as well as digital texts. Educational schemas attempt to operationalise the peculiar textual structures of traditional learning materials and learning conversations, where a learner’s relation to text is configured into an interchange not unlike the ATM conversation we described in Chapter 4, ‘What does the digital do to knowledge making?’. E-commerce and digital rights management schemas move texts around in a world where intellectual property rights regulate their flow and availability.

Tagging schemas and processes may be represented as paradigm (using syntagmatic devices such as taxonomy) or as narrative (an account of the ‘funnel of commitment’ and the alternative activity sequences or navigation paths in the negotiation of that commitment). Ontologies are like theories, except, unlike theories, they do not purport to be hypothetical or amenable to testing; they purport to tell of the world, or at the very least of a part of the world, like it is—in our case that part of the world inhabited by authors, publishers, librarians, bookstore workers and readers.

The next generation of ontology-based markup brings with it the promise of more accurate discovery, machine translation and, eventually, artificial intelligence. A computer really will be able to interpret the difference between Cope, cope and cope. Even in the case of the < author > with the seemingly unambiguous < surname > Kalantzis, there is semantic ambiguity that markup can eliminate or at least reduce, by collocating structurally related data (such as date of birth) to distinguish this Kalantzis from others and by knowing to avoid association with the transliteration of the common noun in Greek, which means ‘tinker’.

In the world of XML, tags such as < author > and < surname > are known as ‘elements’, which may well have specified ‘attributes’; and the ontologies are variously known as, or represented in, ‘schemas’, ‘application profiles’ or the ‘namespaces’ defined by ‘document type definitions’ or DTDs. As our interest in this chapter is essentially semantic, we use the concepts of ‘tag’ and ‘schema’. In any event, as mentioned earlier, ‘ontology’ seems the wrong concept insofar as tag schemas are not realities; they are specific constructions of reality within the frame of reference of highly particularised social languages. Their reality is a social reality. They are no more and no less than a ‘take’ on reality, which reflects and represents a particular set of human interests. These interests are fundamentally to get things done (funnels of commitment) more than they are mere reflections of objectified, inert being. Schemas, in other words, have none of the immutable certainty implied by the word ‘ontology’. Reality does not present itself in an unmediated way. Tagging schemas are mediated means rather than static pictures of reality.

Most of the tagging frameworks relating to authorship and publishing introduced above were either created in XML or have now been expressed in XML. That being the case, you might expect that an era of rapid and flexible transmission of content would quickly dawn. But this has not occurred, or at least not yet, and for two reasons. The first is the fact that, although almost all content created over the past quarter of a century has been digitised, most digital content has been created, and continues to be created, using legacy design and markup frameworks. These frameworks are embedded in software packages that provide tools for working with text which mimic the various trades of the Gutenberg universe: an author may use Word; a desktop publisher or latter-day typesetter may use Quark; and a printer will use a PDF file as if it were a virtual forme or plate. The result is sticky file flow and intrinsic difficulties in version control and digital repository maintenance (Cope 2001). How and where is a small correction made to a book that has already been published? Everything about this relatively simple problem, as it transpires, becomes complex, slow and expensive. However, in a fully comprehensive, integrated XML-founded file flow, things that are slow and expensive today should become easier and cheaper—a small change by an author to the source text could be approved by a publisher so that the very next copy of that book purchased online and printed on demand could include that change. Moreover, even though just about everything available today has been digitised somewhere, in the case of books and other written texts, the digital content remains locked away for fear that it might get out and about without all users paying for it when they should. Not only does this limit access, but what happens, for instance, when all you want is a few pages of a text and you do not want to pay for the whole of the printed version? And what about access for people who are visually impaired? It also puts a dampener on commercial possibilities for multichannel publishing, such as the student or researcher who really has to have a particular text tonight, and will pay for it if they can get it right away in an electronic format— particularly if the cost of immediate access is less than the cost of travelling to the library specially.

The second reason is that a new era of semantic text creation and transmission has not yet arrived. Even though XML is spreading quickly as a universal electronic lingua franca, each of its tagging schema describes its worlds in its own peculiar way. Tags may well be expressed in natural languages—this level of simplicity, openness, transparency is the hallmark of the XML world. But herein lies a trap. There is no particular problem when there is no semantic overlap between schemas. However, as most XML application profiles ground themselves in some ontological basics (such as people, place and time), there is nearly always semantic overlap between schemas. The problem is that, in everyday speech, the same word can mean many things, and XML tags express meaning functions in natural language.

The problem looms larger in the case of specialised social languages. These often develop a high level of technical specificity, and this attaches itself with a particular precision to key words. The more immersed you are in that particular social language—the more critical it is to your livelihood or identity in the world, for instance—the more important these subtle distinctions of meaning are likely to be. Communities of practice identify themselves by the rigorous singularity of purpose and intent within their particular domain of practice, and this is reflected in the relative lack of terminological ambiguity within the social language that characterises that domain. As any social language builds on natural language, there will be massive ambiguities if the looser and more varied world of everyday language is assumed to be homologous with a social language which happens to use some of the same terminology.

The semantic differences between two social languages in substantially overlapping domains are likely to be absolutely critical. Even though they are all talking about text and can with equal facility talk about books, it is the finely differentiated ways of talking about book that make authors, publishers, printers, booksellers and librarians different from each other. Their social language is one of the ways you can tell the difference between one type of person and another. These kinds of difference in social language are often keenly felt and defended. Indeed, they often become the very basis of professional identity.

This problem of semantics is the key dilemma addressed by this chapter, and the focal point of the research endeavour which has produced CGML. Our focus in this research has been the means of creation and communication of textual meaning, of which the book is an archetypical instance. Each of the schemas we have briefly described above channels a peculiar set of ‘funnels of commitment’ in relation to books—variously that of the author, typesetter, printer, publisher, bookseller, librarian and consumer. And although they are all talking about the same stuff—textual meaning in the form of books or journal articles—they talk about it in slightly different ways, and the differences are important. The differences distinguish the one funnel of commitment, employing its own peculiar social language to realise that commitment, from another. It is precisely the differences that give shape and form to the tagging schemas which have been the subject of our investigations.

The schemas we have identified range in size from a few dozen tags to a few thousand, and, the total number of tags across just these schemas would be in the order of tens of thousands. This extent alone would indicate that the full set of tags provides the basis for a near-definitive account of textual meaning. And although it seems as if these schemas were written almost yesterday, they merely rearticulate social languages that have developed through 500 years of working with the characteristic information architectures of mechanically reproduced writing, bibliography and librarianship, the book trade and readership. Given that they are all talking about authorship and publishing, the amount of overlap (the number of tags that represent a common semantic ground across all or most schemas) is unremarkable. What are remarkable are the subtle variations in semantics depending on the particular tagging schema or social language; and these variations can be accounted for in terms of the subtly divergent yet nevertheless all-important funnels of commitment.

So, after half a century of computing and a quarter of a century of the mass digitisation of text, nothing is really changing in the core business of representing the world using the electrical on/off switches of digitisation. The technology is all there, and has been for a while. The half-millennium-long shift is in the underlying logic behind the design of textual meaning. This shift throws up problems which are not at root technical; rather they are semantic. Interoperability of tagging schemas is not a technical problem, or at least it is a problem for which there are relatively straightforward technical solutions. The problem, and its solution, is semantic.

The commercial implications of the emergence and stabilisation of electronic standards are also enormous. These include:

The key issue is the flow of business information and content between the various players in the text production and supply chain. Addressing this issue can produce efficiencies and competitive advantage for individual enterprises, and the whole industry. Many players are now arguing, in fact, that addressing this issue is not a choice—it is a necessity given the fact that standards are rapidly emerging, stabilising and gaining wide acceptance.

CGML is an attempt to address these semantic and commercial challenges. The aim of the CGML research and development endeavour has been to develop software that enables digital text production (electronic renderings and print renderings) using a markup language that offers stable and reliable interoperability across different standards. These standards include typesetting and text capture, electronic rendering, print rendering, B-2-B e-commerce, e-learning, digital rights management, internet resource discovery and library cataloguing.

CGML addresses one of the fundamental issues of the ‘semantic web’— the problem of interoperability between different but overlapping and related electronic standards. Commercially and functionally, the intended result is a software environment in which texts render simultaneously to electronic and print formats (for instance, a bound book, computer screen, handheld reading device or synthesised voice) from a common source file. Metadata generated by this software is simultaneously able to create a library cataloguing record, an e-commerce record (automated entry to Amazon, international bookdata databases etc.) and make a published text an e-learning object and conform to the current and emerging digital rights management protocols.

At the time of writing, CGML consists of approximately 1,000 tags, interpolated into an XML schema. These tags are defined in the Common Ground Dictionary of Authorship and Publishing, which currently runs to some 25,000 words. CGML and the Dictionary are published dynamically (without ‘editions’ or ‘versions’), with tags being constantly added and definitions refined as the Common Ground research endeavour proceeds. The Common Ground research and development team has created its own ontology-building software, CGMeaning, which houses CGML as well as providing a foundation for the export of data into a range of XML text and publishing schemas.

CGML takes the textual artefact of the written text as its point of departure. However, insofar as the digital medium also serves as a construction tool, repository and distributional means for audio, moving image, software, databases and the like, CGML also incorporates reference to these representational forms. CGML is designed to be fully extensible into all domains of creative media and cultural artefact.

From a technology point of view, CGML sets out to tackle one of the fundamental challenges of the ‘semantic web’—the problem of interoperability between overlapping and related electronic standards. To this end, Common Ground researchers developed the ‘interlanguage’ mechanism (Common Ground Publishing 2003). This mechanism has the potential to extend the useability of content across multiple standards, XML schemas, ontologies or database structures. The approach taken by CGML may begin to address the enormous problem of interoperability in general, not just in publishing but in other areas of semantic publishing. Stated simply, electronic files do not flow well along production and distribution supply chains, because file formats vary, as does the metadata which defines file content format and uses. In the case of published material, there are enormous inefficiencies in file flow from author to publisher to printer to electronic rendering formats as well as the e-commerce mechanisms that deliver content from enterprise to enterprise and finally to consumers.

Even though each electronic standard or schema has its own functional purposes, there is a remarkable amount of overlap between these standards. The overlap, however, often involves the use of tags in mutually incompatible ways. Our extensive mapping of 17 standards in various text- and publishing-related fields shows that, on average, each standard shares 70 per cent of its semantic range with neighbouring standards. Despite this, it is simply not possible to transfer data generated in one standard to a database or XML schema using another. Each standard has been designed as its own independent, stand-alone schema. This, in fact, points to one of the key deficiencies of XML as a metamarkup framework: it does not in itself suggest a way for schemas to relate to each other. In fact, the very openness of XML invites a proliferation of schemas, and consequently the problem of interoperability compounds itself.

This produces practical and commercial problems. In the book publishing and manufacturing supply chain, different links in the chain use different standards: typesetters, publishers, booksellers, printers, manufacturers of electronic rendering devices and librarians. This disrupts the digital file flow, hindering supply chain integration and the possibilities of automating key aspects of supply chain, manufacturing and distribution processes. Precisely the same practical problems of interoperability are now arising in other areas of the electronic commerce environment.

Although our main interest is the world of authorship and publishing, the longer term possibilities technologies of interoperability such as CGML are in the areas in which the semantic web has so much—as yet unfulfilled—promise. This includes: indexing, cataloguing and metadata systems; product identification systems; systems for the production, manufacture and distribution of copyright digital content; knowledge and content management systems; systems for multi-channelling content and providing for disability access; machine translation from one natural language to another; and artificial intelligence.

More practically, the challenge of interoperability is this: in a scenario where there are many more than two parties, where the information is not covered by a single standard, where the resources and skills of the parties cannot facilitate costly and time-consuming integration, an approach is needed that caters for the complexity of the messages, while providing tools that simplify the provision and extraction of data and metadata. This is the crux of semantic interoperability. Such an approach involves providing a systematic mapping of associated XML standards to a common XML ‘mesh’, which tracks semantic overlays and gaps, schema versioning, namespace resolution, language and encoding variances, and which provides a comprehensive set of rules covering the data transfer—including security, transactional and messaging issues.

The idea of a ‘meta-schema’—a schema to connect related schemas— was initially considered to be sufficient. Research has demonstrated, however, that this is not enough, being subject to many of the same problems as the individual schemas being mapped—versioning, terminological differences and so on.

The core operational principles of CGML are as follows: meaning form or rendering is rigorously separated from, yet reliably follows, markup tags expressing meaning function; interoperability of tagging schemas can be achieved by mapping through an interlanguage governed by a set of semantic and structural rules; a tag schema expresses paradigmatic relations; a tag thesaurus expresses relations between tagging schemas; a tag dictionary expresses semantics; interoperability mechanisms are automated or semi-automated; and tag narratives anticipate a range of activity sequences driven by funnels of commitment and realised through alternative navigation paths.

As a terminological aside, CGML deliberately recruits some quite ordinary concepts from the world of textual meaning, such as the ideas of a dictionary, thesaurus and narrative. If we are going to have what we have termed ‘text-made text’ in which markup is integral to textual reproduction, we might as well use these historically familiar devices—albeit with some refinement and levels of precision required by the logistics of digital meaning.

Interlanguage

Semantic publishing schemas promise to overcome two of the most serious limitations of the World Wide Web: searching involves simply identifying semantically undifferentiated strings of characters, and rendering alternatives is mostly limited by data entry methods—printed web pages do not live up to the historical standards of design and readability of printed text, and alternative non-visual renderings, such as digital talking books, are at best poor.

Specific schemas are designed to provide more accurate search results than is the case with computer or web-based search engines. Examples include the Dublin Core Metadata Framework and MARC electronic library cataloguing system. However, metadata harvested in one scheme cannot be readily or effectively used in another.

Specific schemas are also designed for a particular rendering option. For instance, among schemas describing the structure of textual content, HTML is designed for use in web browsers, DocBook for the production of printed books, Open eBook for rendering to handheld reading devices and Digital Talking Book for voice synthesis. Very limited interoperability is available between these different schemas for the structure of textual data, and only then if it has been designed into the schema and its associated presentational stylesheets. Furthermore, it is not practically possible to harvest accurate metadata from data, as data structuring schemas and schemas for metadata are mutually exclusive.

The field of semantic publishing attempts to improve on the inherent deficiencies in current digital technologies in the areas of resource discovery (metadata-based search functions) and rendering (defining structure and semantics in order to be able to support, via stylesheet transformations, alternative rendering options).

In support of semantic publishing, CGML attempts to inter-relate the principal extant in the tag schemas for the world of authorship and of publishing. However, unlike other tag schemas in this domain, it does not purport to be ontologically grounded. It does not attempt to name or rename the world. Rather, CGML builds a common ground between contiguous and overlapping tag schemas which already purport to name the world of authorship and publishing. It is not a language. It is an interlanguage (Figure 13.2).

The challenge of interoperability of tagging schemas (standards, application profiles or namespaces) has typically been addressed through schema-to-schema ‘crosswalks’. A crosswalk is a listing of tag-to-tag translations not dissimilar from a language-to-language dictionary. For instance, as mentioned earlier, crosswalks have been created between MARC and ONIX (MARC Standards Office 2000) and between MARC and Dublin Core (MARC Standards Office 2001). As Paskin notes, when there are N schemas (N/2)(N–1), mappings are required (Paskin 2003). For instance, as of writing, CGML maps to 17 schemas. For full interoperability, 136 ‘crosswalk’ mappings would be required. Or, to take a natural language analogy, if there are 60 languages in Europe, translation between all 60 languages can be achieved with 1,770 language-to-language dictionaries—Italian–Gaelic, Gaelic–Vlach, Vlach–Italian and so on.

In fact, things are more complicated even than this. Each dictionary is, in fact, two dictionaries. Italian–Gaelic and Gaelic–Italian are not mirror inversions of each other because each language frames the world in its own semantically peculiar way. Similarly, the MARC to ONIX exercise (MARC Standards Office 2000) is quite a different one from the ONIX to MARC exercise (EDItEUR 2003). MARC to ONIX translates a library cataloguer’s understanding of the nature and content of the book into a form intelligible to a publisher or a bookseller; and ONIX to MARC translates a publisher’s or bookseller’s understanding of the book into a form intelligible to a library cataloguer. In each case, the frame of reference or the starting point is defined in terms of a subtly distinctive social language. Each crosswalk is a separate intellectual and discursive exercise. So, we need to modify Paskin’s crosswalk formula as follows: the number of mappings to achieve interoperability between N tagging schemas is 2[(N/2)(N–1)]. In a terrain encompassed by the current version of CGML, 272 crosswalks would be required; Europe needs 3,540 dictionaries for comprehensive cross-translation of all its languages. (And, while we are on this train of thought and although it is tangential to our point, cross-translation of all the world’s estimated 6,000 languages would require a practically impossible 17,997,000 dictionaries.)

Creating a single crosswalk is a large and complex task. As a consequence, the sheer number of significant overlapping tagging schemas in a domain such as authorship and publishing presents a barrier to achieving interoperability—and this without taking into account the fact that the schemas are all in a state of continuous development. Every related crosswalk needs to be reworked with each new version of a single tagging schema. Moreover, new tagging schemas are regularly emerging and every new schema increases the scale of the problem exponentially. Five cross-translations require ten crosswalks; ten cross-translations require 90 crosswalks.

Paskin suggests that this level of complexity can be eased by mapping ‘through a central point or dictionary’ (Paskin 2003). This is precisely the objective of CGML, which is an intermediating language, or an interlanguage through which a full set of translations can be achieved. Tag by tag, it represents a common ground between tagging schemas. Tag < x > in the tagging schema A translates into tag < q > in CGML, and this in turn may be represented by < y > in tagging schema B and < z > in tagging schema C. The ‘common ground’ tag < q > tells us that < x >, < y > and < z > are synonyms. A theoretical 272 crosswalks are replaced by 17 thesauri of tag synonyms (Figure 13.3). If, by analogy, all European languages were to be translated through Esperanto, a language deliberately fabricated as a common ground language, 60 dictionaries would be needed to perform all possible translation functions instead of a theoretical 3,540. Even simpler, in theory just one dictionary would suffice, translated 60 times with 60 language-to-Esperanto thesauri. This is precisely what CGML does. It attempts to solve the semantic aspect of the interoperability problem by creating one dictionary and 17 thesauri of tag synonyms. (And, incidentally, returning to natural language for a moment, this technique can be used as a semantic basis for machine translation, bringing the inter-translatability of all human languages at least into the realm of possibility.)

Figure 13.4 shows the interlanguage approach—full interoperability between 17 schemas requires a thesaurus with just 17 sets of tag synonyms.

CGML has a number of distinguishing features, which means that it is constitutively a very different kind of tagging schema from all the others against which it maps. It is this constitutive character that defines it as an interlanguage, as distinct from a language.

An interlanguage has no life of its own, no independent existence, no relation to reality other than a mediated relationship through other languages. We will outline the operational principles for the construction of such an interlanguage through the subsequent subsections of this chapter.

Before this, however, we want to mention some of the unique characteristics of an interlanguage such as CGML. As an interlanguage, CGML is designed to be open to the possibility of mapping new schemas that may emerge within or substantially overlapping its general domain. It is also designed to be able to absorb tagging that finely distinguishes conceptual subsets of certain of its core interests. In the case of authorship and publishing this might include, for instance, geospatial tags to define precise location, or tags representing controlled subject vocabularies in specific field-domains. By comparison with the crosswalk alternative, this mapping is achieved with relative ease.

Full subsumption and overlap are both cases of vertical integration of tagging schemas into CGML. However, CGML is also designed to be amenable to horizontal integration of schemas defining contiguous or complementary domains, such as the integration of other digital media or museum objects with the world of books. After all, books are routinely made into movies, bookstores sell DVDs and printed books and libraries store individual copies of rare and unique books as artefacts.

As an interlanguage, CGML is infinitely extensible, absorbing subsidiary, overlapping and contiguous schemas to the extent that seems necessary and useful. A the time of writing, CGML consists of nearly 1,000 tags—and these happen to be the tags for which there is the greatest degree of semantic common ground identifiable as synonyms across the interlanguage. The tags that represent the greatest degree of overlap also happen to be the most fundamental to the representational and communicative activities of authorship and publishing. However, there is no reason why CGML should not extend to 10,000 or a 100,000 tags as it describes progressively more arcane bywaters in each tagging domain (vertical integration) or as it spreads its range of reference into contiguous domains of meaning (horizontal integration).

To reiterate, CGML is an interlanguage which maps against any other related schema (or, as they are variously termed, standards, namespaces or application profiles) in the domain of authorship and publishing. It works through tag-to-tag translation links between schemas—be they competing within a substantially overlapping domain or serving varied functions in divergent but still overlapping domains. The CGML term is an intermediary or interlanguage term. CGML is not a schema in and for itself. Rather, it is a way of talking to schemas.

The conventional approach to evaluating the efficacy of alternative tag schemas within a particular semantic domain is to undertake a process of comparison and contrast, the purpose of which is to select the one that would, it seems for the moment at least, be most appropriate to one’s expressive needs, or the one that appears to be the most internally coherent and robust, or the one that happens to be most widely used among the players within a particular community of practice.

As an interlanguage, however, CGML is entirely agnostic about the ontological validity of the schemas to which it maps. If they move a community of practice, or have the potential to move a community of practice, they are worth the trouble of mapping. New standards may emerge, and if they appear to be sufficiently cogent and practically useful, they are also worth the trouble.

CGML itself does not set out to be a competing or alternative standard. Rather CGML takes the approach that the prevailing uncertainty about which standards will predominate and the likelihood of the emergence of new standards is to a significant degree a diversion. In the interlanguage approach, standards are everything—CGML needs to talk with the main existing and emerging publishing standards from the pragmatic point of view of interoperability. Yet, in another sense, standards are nothing—it is immaterial if some standards fall into desuetude or if new standards emerge. Dogmatic debate about the value or lack of value of a particular schema or standard is of little value. Shoehorning social practices into ill-fitting received standards is also a fraught exercise. CGML cares about standards but eschews standardisation, or making things the same for the sake of sameness.

Our decision to take the interlanguage approach, paradoxically in the light of our scepticism about the ontological pretensions of tag schemas, is based on the stability inherent to the semantic ground, or a kind of ontological pragmatism. Behind the varied ‘takes’ on reality reflected by tag schemas, there is still a relatively stable and thus predictable material and social reality. The ‘resistances’ of which Eco speaks are at times insistent. Although we conceptualise the world paradigmatically through tag schemas and operationalise these schemas through activity narratives, these paradigms and narratives do have a reference point, and this reference point might reasonably be construed to be a matter of ontology. Ontology does not simply present itself; it is mediated by paradigms and narratives. However, ontology practically grounds paradigm and narrative. In fact, through language, paradigm and narrative make themselves integral to the ontological reality of society or culture.

This grounding provides stability and thus a certain predictability of paradigm and narrative within a particular semantic domain. If authorship and publishing is our domain of interest, for instance, this represents a set of social practices—practices of representation and communication—that have deep and only gradually changing roots. There are authors who write; these authors have names; their writings have titles; and these writings have characteristic generic structures and fields of representation or subjects. Any new tagging schema that turns up—no matter how fancy it is or how innovative its intentions and methodology (e-learning, digital rights management, variable rendering and the like)—is still going to have to name these insistent realities.

The basis of CGML, in other words, is in the semantic ground of publishing, and there is an essential stability in the everyday lifeworld of authorship and publishing. The technologies may be changing, but there are still creators (writers, editors, illustrators) creating works (books in print and electronic formats, chapters, articles and other written, visual and audio texts) that are subject to copyright agreements, which are then used by consumers (readers, learners). Schemas do no more than represent that lifeworld from a particular perspective—be that the perspective of the library, digital resource discovery, rights, commerce, education or rendering/production. Schemas may come and go, but the lifeworld they purport to represent and facilitate remains relatively stable. At most, it changes incrementally, despite even the large changes underway today in the new digital media.

The interlanguage approach of CGML also provides a tool for literature, science and curricula to be built in small languages and endangered languages, including, with the aid of Unicode, publication in any script. CGML can run in any language and any script, and this is achieved simply by translating the tags and tag definitions. This may seem a relatively small move in a practical sense. Conceptually, however, it is a huge move. In fact, it turns a linguistically expressed term into a mere ‘token’ of a core concept that exists above and beyond any particular language. And an indirect effect of this move is to add multilingual functionality to markup schemas which currently exist only in English. In addition, by virtue of its structural and semantic approach to markup, CGML could serve as an aid to effective and accurate human and machine translation. In other words, by these various means, CGML could literally find itself in the space of an interlanguage between various human languages.

In a globalised and multilingual world, Ron Scollon argues, social languages or discourses are more similar across languages than within languages (Scollon 1999). The way academics write for their particular discipline, for instance, whether it is in English or Japanese, is similar in terms of the structure of their texts and the ways those texts describe the world. A structural and semantic framework for structuring text such as CGML, which includes elaborate structural and semantic markup linked to controlled keyword vocabularies, will work across languages once the tags and the specialist vocabularies are translated, and this is because the most important thing about the discourse does not sit inside a particular language. Text structured and rendered in this way may become the platform for multilingual, multi-script publishing in communities more and more defined by their social language (what they want to do in the world, as expressed in peculiar ways of communicating about the world) than by the accident of mother tongue.

The CGML Dictionary does not purport to be about external referents as ‘meaning’; rather, it is built via the interlanguage technique from other languages which purport to have external referents. Moreover, insofar as the semantic ground of CGML is meaning itself (and its instantiation in the practices of authorship and publishing), it is a kind of metasemantics, a language of meaning. It happens to be centred on the realm of semantics in general—the meaning of meaning—and within that realm the social practices and technologies of representation and communication stabilised in the historical practices of representation.

Furthermore, CGML is not an ordinary dictionary insofar as it develops a specialised ‘take’ on the world it purports to describe, the world of meaning. Its meanings are not the commonsense meanings of the lifeworld of everyday experience, but derivative of specialised social languages which speak in the refined and particularistic way characteristic of the professionals and aficionados of that domain. To apply a pair of concepts of Husserl’s, commonsense language is shifting and ambiguous language of the lifeworld; social languages develop progressively refined (sedimented) and self-consciously reflective (bracketed) discourse more characteristic of science (Cope and Kalantzis 2000; Husserl 1970). CGML, in other words, derives from schemas developed in and for professions which have developed high levels of conceptual clarity about what authorship is and what publishing involves.

The CGML dictionary links a notation (the tag-concept), which may be used in practice as a label for a field in a database or as an XML tag, to a semantically explicit definition, as well as an annotation which explains and exemplifies the tag-concept in terms of subordinate tag-concepts in the taxonomy (the various logics of relation-inclusion discussed earlier), and provides advice where necessary on appropriate and well-formed data entry. The building blocks of the CGML dictionary are the other tag-concepts of the CGML schema, and these are connected by hyperlinks. The definition builds on parent tag-concepts; the annotation suggests the possible instantiations of a tag-concept by means of illustrative child tag-concepts. The dictionary is located in maintained in a purpose-build software application, CGMeaning. Figure 13.5 shows how it specifies the concepts of < creation > and < creator > .

References