Do not copy, quote, or redistribute without permission of the author
A decrease of user interaction in the drift to "push" is part of a search for mechanisms of identity and credibility on the web. Users typically bring only the evaluative processes they apply to "traditional" texts. Better understanding and implementation of hypertexts are needed to prevent the web from becoming television delivered in packets. Dictionaries show how quality and authority can be achieved in hypertext. They are both "traditional" documents as well as canonical hypertexts: hypertext is their core mechanism for revealing meaning to users. There are several implications: hypertext is a technology of text and literacy, not a software technology; we must regard hypertexts as resources that support users' hypertext activity, not merely providing hypertext links; some genres of texts can support users' hypertext activities directly rather than merely implement hypertext through a contrived interface; and finally centripetal (ie inward-pointing) hypertexts are a locus for authority on the web.
The web's most pressing current problem is to find ways for representing identity and quality. There are insufficient mechanisms for users to differentiate hypertext documents, and to detect and assign levels of trust and authority to them.
Comparing the web's state to a crossroads leading one way to global, democratic information sharing, the other way to total mediocrity, Ciolek (1996) summarised strategies to rescue the web from imminent quality meltdown. These strategies consisted of enriching the presentation interface (Ciolek called this a "programming approach"), publishing explicit rules or conventions for web content ( "procedural approach"), meta-data systems ( "structuring approach"), enriched naming and indexing methods ("bibliographical approach"), review publishing based on quality parameters for evaluating content ( "evaluative approach"), and finally complementing web content with supporting social/professional networks ( "organisational approaches").
This paper complements Ciolek's typology by describing "quality" in relation to the semiotics of the web's distinguishing feature - hypertext, or links between texts. Quality cannot be imposed, or even described; it arises from activations of conventions that are shared between producers and users. Increasing or even maintaining the quality of the web's resources will depend on better understanding these semiotics of links. Dictionaries - both traditional printed ones and web versions - have a special role in the hypertext universe because they offer an explanation of how authority and quality arises naturally within hypertext media.
a complete system of reference to primary sources of knowledge ... growing and changing continually ... [like a] nervous network, a system of mental control about the globe, knitting all the intellectual workers of the world through a ... common medium of expression into a more and more conscious and co-operating unity(Wells 1971, quoted in Franklin & Kinnell 1990:24).
Theodore Nelson coined the name "hypertext" in the 1960's as new computer technologies made possible the implementation of systems proposed by Wells and Bush.
Now is the time to remind ourselves of hypertext's ancestry and refocus on its defining feature - the implementation of links. You do not need a machine to find hypertext; it has long been with us before its links were electronically enabled. Many written texts, especially scholarly or authoritative ones, are routinely read in a hypertextual, non-sequential fashion. The reader departs from the text's flow to check a footnote, reference, or index entry, reads it fully or partially (perhaps extracting and recording some of the information there), before returning to the original place or a location suggested by the newly accessed information (Landow 1989:174).
The web's implementation has departed from the visions of its early visionaries in one significant way. Bush envisioned that users would freely create and share "associative trails" amongst documents: "any item can be caused at will to select immediately and automatically another ... The process of tying two things together is the important thing" (quoted in Franklin & Kinnell 1990:25). However, on the web, while the "associative trails" (hypertext links) do implement sharing, the average user's ability to create shared associations is relatively limited.
The producer-user or reader-writer relationship has turned out to be quite different under the real operating conditions of the Internet, where the two processes of creating and using links are far from symmetrical, due to the complexity of access to web serving and the sheer amount of work required for web publishing (Ciolek 1995, Green 1996).
Despite increasing access to web publishing in many developed countries, the ratio of users to publishers remains comparable to the traditional paper publishing where many read and few publish. Rhetoric about the universality of the web has not been complemented by mechanisms to help us transcend traditional notions (and practices) of writer and reader. Today's published hypertext web resources are classified and evaluated by the public as the work of writers rather than as the tentative creations of Bush's reader/writers because users have only the tools supplied by their societies' existing literary practices (which are in many cases highly developed and deeply-embedded, but see also Nathan 1997).
As a result we find ourselves in the paradoxical situation that despite web publishing remaining more centralised than its original design, concern is rapidly increasing not only about the quality of material available on the web but our ability to distinguish the valuable from the dubious. This is crucial in areas such as education:
it is important that users of information from the Web verify the quality of information ... For example, using a search engine primarily returns 'soft' information whose quality is questionable. On the other hand, the Web also allows access to 'hard' information ... The implication is that teachers and students need to be able to differentiate between the hard and the soft information(Atkinson 1996:13.)
One way to approach this paradox is to analyse the core properties of hypertext through its canonical and oldest exponent - the dictionary. Dictionaries create authority through their pattern of hypertext. The next sections of this paper discusses aspects of dictionary hypertextuality that shed light on the nature of hypertext and how authority and trust can arise as an inherent property.
Firstly, note that no one actually reads a dictionary: there is no narrative structure, no plot, or even beginning or end (because the listing order is typically the arbitrary alphabetic one). Indeed, their internal cross- references often make them circular to use.
Dictionaries have a far greater "density of links in relation to ... content" than other books (Woodhead 1991: 143). Ordinary grammatical structures are suppressed. Words are isolated and abbreviated. Phrasing is clipped and telegraphic, relying on the context of use to reveal meaning. The defining vocabulary may even be drawn from a fixed set (for example, the Longmans Dictionary of Contemporary English). Rarely is reference made to real things; dictionary definitions use indefinite articles.
In other words, dictionaries are texts that do not evoke language, but instead represent metalinguistic meaning through minimal linguistic stock but enriched textual devices; alphabetical systems, typographical devices and spacing, special codes, and cross referencing.
The information and visual structures of dictionaries' supports their lookup function. You read a dictionary as a detour: it is an exterior text that leads you to the dictionary, from where you may return to that text. The structure of its readings is variable, depending on the interaction between a user's task in relation to an exterior text and her background knowledge, and the information access paths provided by the lexicographer. Their structures enable readers to have a large number of internal, cross-referencing links, and maximal access points into this network.
A brief encounter with dictionaries of Kanji languages (Chinese, Korean, and Japanese) will confirm the extent to which dictionary hypertext relies on its underlying text technology. To look up a word in a Kanji dictionary you begin with the shape of the word character, and then follow a chain of lookups depending on the number of subcharacters (radicals), or number of strokes. Recognition and choice amongst sets of characters finally converge to an item within a group with related meanings. While the Kanji dictionary's external function is the same as alphabetical dictionaries, the lookup process is radically different because the writing system is.
Intertextuality theory (cf Snyder 1996, Nathan 1996) takes a dynamic view of language and text where texts (including spoken texts) take part in an ongoing tension with the language "system" to both create meaning and modify the meaning potential of language. Language itself is not so much a system as the cumulative result of text productions. All texts are thus interdependent through their commerce with language.
When people read, they produce meaning, not solely directed by the text they read but also indirectly through their knowledge of other texts.
Dictionaries are intertextual in two ways. Firstly, they have "outer" intertextual relations: they redeem the exterior text that baffles a reader by listing and relating meanings. Dictionaries seem at least to list complete vocabularies and so potentially intersect with every text of the language: in a sense all texts lead to the dictionary. Secondly, they have the complex system of inner, centripetal typographical devices described earlier.
The third layer - self-construction - results from the resemblance of the concrete centripetal network to the potential network expressed by intertextual relations. This allows dictionaries to both appear and function as a microcosm of the whole field of texts of their language.
Dictionary cross-references are the valid set of paths or links in the dictionary space, imposed by the lexicographer as a filter on the potentiality of links between words. The link relationships map the lexicographer's expert knowledge (Jonassen, D. 1990) of the language's semantics using a visual format whose conventions the user shares. Because the actual hypertext devices used in dictionaries correspond to actions that users find familiar and predictable (such as looking up a synonym), readers participate in rational processes of hypertext arrival and departure (Landow 1986).
Note the redistribution of knowledge and authority. While lexicographers give up much of an author's usual authority (because they do not force any two readers to read the same words), and readers have an unparalleled choice of paths (and also of what to ignore: see below) and point of departure, the result is a genre where readers can experience the author's expertise through hypertext itself. And since in the case of dictionaries, centripetal hypertext is a source of self-constructed authority, user freedom correlates with evaluation of the text as authoritative.
A corollary of visual attention is visual attenuation. This is used frequently in dictionaries by using conventional typographic devices to suppress segments of the text. For example, most readers know that if they ignore the italic abbreviations that follow a headword (often the grammatical or phonetic information), their consultation of the dictionary is not rendered invalid. Similarly, the list notation used to present the different senses of words allows users to selectively chunk, pay attention to, ignore, or choose.
The tension between providing rich information environments, with differentiated parts to allow users license to select items of interest and action can be called graded context. On one hand, "graded context" reflects the contrast between hypertext and form-based interfaces to databases that are not true dictionaries (see below). It also allows us to mobilise the new technologies available to us to solve a problem for screen-based dictionaries.
Screen-based dictionaries can suffer because entries can run off the screen and make the search for information more difficult. Ultimately, this problem denies the user benefits of the hypertext properties of the dictionary genre itself, and it is paradoxical that some electronic dictionaries find themselves throwing away some of their precomputational hypertextual properties in their on-line versions.
We are currently creating new types of dynamic-display dictionaries that expand on the notion of graded context. Users will see the core structures of the dictionary, with all its essential elements viewable in context, or choose to enlarge the scope of individual entries. A dictionary implementing this strategy (A Beginners' Dynamic Dictionary of Telugu) is currently under development and will be available on the web at <http://www3.aa.tufs.ac.jp/~bhaskar/teldict.htm> before the WWW7 conference.
But the strategy also has wider implications for the web. "Graded context", or, put another way, mechanisms for variably suppressing visual information are equally important textual devices for achieving hypertext properties as the actual links between chunks of text. In some ways, the subtlety in which suppression can be implemented, as illustrated by the case of dictionaries, suggests that this is an area for much greater elaboration and understanding before hypertext media is mature.
Underlying such resources is an assumption that the dictionary entry is the interface and therefore the text. While still enormously useful, this makes for a limited resource because the interface does not offer the granularity and choice of devices that users need.
While fast and powerful for many tasks, this kind of on-line dictionary does not deliver functions that even the traditional printed dictionary genre fulfils. It does not display the scope of the dictionary, or indeed the language itself; it does not provide opportunity for heuristic or even accidental discovery; it demands that the user know at the outset what she probably has least knowledge of - the correct spellings for target language words. As a result, this kind of interface often fails a user, by returning the result "Your query did not match."
Compare a dictionary that supports all the "traditional" hypertext, such as The Kamilaroi/Gamilaraay Dictionary. This dictionary offers users the ability to browse, choose, and make interpretations at various levels of structure. The contrast between these two approaches to on-line dictionaries is the distinction between true hypertext systems that support users' hypertextual activities, versus systems that implement linked data structures.
More generally, it is worth noting the ability of dictionaries to present users with a direct, unmediated interface to the information they want. They already know everything about what they can interact with: in a sense there is no interface at all, only a dictionary. Our screens seem to be occupied by more and more buttons and contrivances fighting for our attention, but ultimately forcing us to interact with application interfaces rather than the materials of interest.
There is a link between the centripetal uses of hypertext described and advocated here, and the expansion of "push" technology. Several contemporary commentators have expressed the fear that the expansion of "push" technology will transform the web into TV (see, for example, Mandl 1997). Push technology decreases the opportunities for the user to interact; interaction is largely reduced to selecting whether or not to receive the offerings within a particular "channel".
The channels themselves come under two new pressures not felt by "traditional" web publishing. The first of these involves production values. When users' interactions are minimised, publishers need to find genres that "make sense" without much user input. Fortunately for such publishers, there are ample existing standard genres, all of which involve some kind of narrative flow (including the "micro-narrative" of the news-story). To impress the users with narrative genres that they are already familiar with, publishers cannot make do with simple web pages: high production values (ie high visible cost, not necessarily quality) are required.
This amounts to merely a slight shift from the television model, where high production values go hand in hand with concentration of publishing in the hands of the few who can afford to assemble large teams and infrastructure. User may end up having to choose from the small set of providers who are able to survive in the high-resource environment.
The second pressure is a corollary of the first: establishing identity on the provider channel. If the providers are to have stability - to attract or generate income - they will need to create or inherit identities that connote authority, class, currency, or other qualities.
Push technology is likely to exploit our society's need for mechanisms to attribute quality and authority. It is already mobilising our desire for trusted sources in order to reduce our opportunities to interact.
The second factor in my pessimism about realising the potential of hypertext is an apparent lack of interest in hypertext among mainstream language or communication theorists. While the late 80's saw significant research in hypertext, continued later with lesser intensity, it is ironical that the pace has slowed down, rather than sped up now that we have real and widely influential application of hypertext to everyday communication.
For example, Roy Harris' recent volume Signs Of Writing did not deem hypertext or even computers worth a mention. Aitchison's (1997) promising sounding title The Language Web similarly has no mention of hypertext (despite including a chapter "The world wide web"!).
Hypertext-supporting features - typically the link/target pair - are perfect and striking examples of written "discourse markers" that present us with a new horizon for investigation because they take the notion of deferred negotiation between producer/receiver (Nystrand 1986) to new limits. However, recent interest in "electronic discourse" has remained confined to descriptions of the pragmatics of interaction; considering the interaction only as the linguistic record or residue left after all the electronic tools have been used. There is no consideration, for example, in Davis and Brewer's Electronic Discourse of the effect of hypertext or electronic navigation systems on communication.
Strangely, current discourse research typically focuses on the pragmatics of exchanges in MUDS and MOOS and newgroups ("discussion forums"). These forums for on-line communication host exchanges where participants have relatively symmetrical roles, and therefore do not offer a context that is either paradigmatically new or typical of the web. Discussions are (merely!) networks of people, not networks of documents. While discussion phenomena are worthy of interest, the hypertext features that occur in them will typically be implemented through the tools of the first technology, human language. Indeed, this is what Coppock (1997:250), for example, shows in his analysis of participants' reference to "virtual events", using various referring expressions of English. While noting that the web is the "most well-known example" of a distributed communication forum, he offers us only the obvious observation that discussion forum exchanges are written yet "seem closer to spoken language than written language".
Web dictionaries can contribute to literacy levels by making hypertext explicit. The explicitness of the web dictionary hypertext has two benefits: by encouraging cross-referencing and browsing, it may help prevent users being misled that there are simple one to one mappings between words in a bilingual dictionary (cf Barratt 1997). A related benefit (not restricted to dictionaries) is that by making hypertext explicit, we can contribute to enhancing literacy skills. In conventional paper-based materials, there is a correlation between the assumed literacy level of the intended audience, and the usage of hypertext. For example, footnotes, abstracts, and indexes of various kinds have in the past been associated with academic writing (see Nathan 1997).
Web dictionaries reveal directions for hypertext because they replicate users' intertextual actions that are the actual mechanism for transacting meaning in book dictionaries (Nathan 1996). As a result they have an inherent capability to present authority and quality to the user, because the user experiences and evaluates that authority while undertaking the meaning negotiation process supported by the dictionary's hypertext mechanisms.
Recently, Beeler (1997) expressed the fear that
Web teaching will replace the ... experience of education in a scholarly atmosphere for a superficial pseudo-learning... If education becomes divorced from its traditional institutions what will prevent it from degenerating into the product of mail-order information factories?The remedy to Beeler's fears lies in quickly establishing widely shared conventions for evaluating information products and information relationships. Ciolek's typology will be found a valuable aid (1996). In addition, we need to understand the actual mechanisms by which sensations of quality are evoked. It has been shown here that dictionaries are a special kind of product that encodes dense relationships between authors, publishers, and users. In turn, that special relationship (and the hypertext principles that underlie it) potentially support very high degrees of quality and authority. (Universities have long been happily placing commercially produced dictionaries at focal locations in their libraries.)
I regard the demonstration of the hypertext properties of dictionaries as a kind of beachhead against the trend to "push". Dictionary users know how to evaluate authority and value through avenues both conventional ("who is the publisher?") and embedded in the experience itself ("was it an enriching experience?", "did it answer my question?"). But most importantly, the case of web hypertext dictionaries demonstrates that hypertext's locus of authority is centripetal, where we can best juxtapose the quality of the content with users' high degree of interaction. Further research is needed to investigate where the boundaries of centripetal hypertext lie.
Users may turn out to be more sensitive to the web's genres than to its technology; and if so the current move to "push", may not succeed in occupying the web-space, and may be increasingly seen by users as simply just another genre, perhaps rather like late-night TV infomercials. Let's hope so; for the future of lexicography, at least.