& 
.
We consider glossaries to be a special form of annotation. It is our belief that annotation is one way that readers can, in a sense, become the joint author of a new work the meaning of which depends, more than usually, on the reader. We have more to say about the similarities and differences between glossaries and annotations in which accompanies the main text.
In the text below,
We use published
rather than shared
to avoid
confusion with the well-established hypertext systems concept [Garg & Scacchi, 1987].
Although annotations can be shared with others (e.g., Blustein's [1996] lists of annotation for a textbook) and are sometimes printed in special editions (e.g., The Annotated Alice [Carroll, 1999]) they are a different class of writing. We term such annotations to be published in the sense that they have been put into a form that is intended for other readers to understand.
Saul's The Doubter's Companion [1995] is an example of a
published glossary.
It is a collection of brief essays about various topics with
short titles arranged alphabetically, in the style of Ambrose Bierce's
famous Devil's Dictionary.
Some of Saul's essays occupy several pages of the printed book.
On the other hand, we use the adjective personal annotation to describe works which are more like the jottings of a moment and not formally arranged for presentation to other readers.
We consider two types of personal glossary using examples from traditional media:
Tied personal glossaries, like the sort we report on in this article, may be most useful for textbooks and similar works. We can imagine a future where electronic books come equipped with glossaries with many entries drawn from the text, much as some printed books do today. The primary difference between our imagined future and today's books is that in the future all of the glossaries will be user-editable (and we hope compatible with each other so that a student can amass a grand floating glossary and compare different definitions of the same term).
Some analogies may help to make the differences clearer. The page in a student's paper notebook where he (or she) records definitions of unfamiliar terms encountered in many books and classes is a form of floating glossary. If the student does not intend to allow other pupils (or even their teacher) to use the glossary, then it is a personal glossary. Similarly, if the teacher provides students with a list of terms and definitions to which the student adds (new terms, definitions, or both) but intends the new or updated entries to be for the student's use only then it is also a personal glossary. The flyleaf of a textbook in which the same student records definitions that pertain only to that book or class is tied because the glossary is connected to a single document. A glossary function provided by a software (e.g. a website or electronic textbook) without the ability to be exported to other programs is similarly tied.
The preceding examples have only dealt with personal glossaries. Shared glossaries are intended for users who might not be the authors of the entries. As such, we can consider that traditional printed dictionaries are a form of glossary. Furthermore if a wiki may be considered to be a glossary (because its structure is of terms and definitions) then it would be a floating one, since it could be accessible to someone who was reading another document, a webpage for example. It could be shared glossary if and only if there were no restrictions on who could read or update the entries.
Differences between glossary types are summarized in the Table below.
published, immediately above)
If our supposition that personal glossaries can help individuals (and co-operating groups of people) to make sense of texts, but that are unsuited for sharing with others, then there is a great opportunity for such tools in ebooks and web browsers. The indicate that our prototypes are good models for tied versions of such products. There is clearly much need for further research and product development in this area.
The main difference between floating and tied glossaries, other than the obvious implementation details, is that the definitions of terms in a tied glossary are not likely to change within the work that the glossary is tied to [Furnas et al., 1987]. We expect that the use of floating glossaries will lead to users encountering conflicting definitions of terms in different documents from within the same community of discourse (as illustrated in the example of ) and of boundary objects in largely unrelated communities [Muller & Friedman, 2000].
The floating tools will clearly be more powerful but they
may, as a consequence, require users to exercise restraint
in their use lest they become unmanageable — think of
how few of the infinity of possible hypertextual structures
are feasible for discursive websites now that the WWW has settled into
genres.
The tactics promoted by Brown & Brown [2004] (namely,
cultivating a community of practice and developing the techniques
as the sophistication and needs of the users grows) might be
appropriate starting point for the development of a floating
glossary tool.
So-called lowercase semantic web
efforts [Çelik & Marks, 2004] are using similar
techniques with apparent success.
(The XHTML Friends Network (XFN) is an example of a lowercase semantic web effort.)
There are however more striking differences between shared and personal glossaries.
We contend that as with shared annotation, making shared glossaries useful is more a matter of human factors than technological sophistication. This view is supported by previous research about glossaries . Also Blustein [2000] found that user interface factors had a greater impact on users' success with a hypertext linking system than the accuracy of the links.
As people read and make notes they are attempting to make sense of a text and this sense-making process alters their interpretation of the text [Dillon, 1994; McKendree et al., 1995; Tague-Sutcliffe, 1995, pp. 8–9, 12–13]. Personal annotations, of which glossary entries are a distinguished type, are the most obvious remnant of the cognitive state that brought them forth. Because it is difficult to reconstruct the meaning of such entries without access to the context in which they were created glossaries that are not carefully and specifically constructed become increasingly difficult for anyone but their author (or compiler) to understand as the number of entries grows.
We speculate that personal glossaries are more suitable than shared glossaries for many tasks. It is well-known that readers rarely agree on where links should begin and end within a single document [Furner et al., 1999; Blustein, 2000]. It is likely that this is due to to way people cognitively process documents [McKendree et al., 1995] and for the same reasons we expect personal annotation to vary amongst individuals. Glossaries in particular are, we believe, an artifact of the reader's cognitive state at the time the definition was recorded. We know that processes of reading and searching for information changes users' cognitive state [Tague-Sutcliffe, 1995] and just as we would not expect a student who had not read a textbook to be able to make sense of a random passage in that book out of context we would not necessarily expect anyone to understand someone else's personal (that is to say, informal) glossary entries.
These differences become clearer in practice.
Marshall &
Brush [2004] conducted a study comparing personal annotations
on paper with shared annotations in a computer network-based forum.
They have found … that annotators make profound changes
to annotations that they share
[Marshall & Brush, 2004, p. 356].
A glossary maintained by a reader in a traditional (codex) text may be kept on a loose piece of paper that the reader moves from page to page as they read — indeed this is often done with mathematical texts when notation is unfamiliar to the reader. If such a glossary includes notes about the relevance of terms and concepts and not merely definitions then it can act as a type of bridge to connect (in the reader's mind) disparate parts of the text. The bridge concept is more similar to an open-linking service [Carr et al., 1998] than to a user-created hypertext link, because the bridge is not explicitly recorded as being between two parts of the text. Since glossaries can provide a way to help readers to make mental associations between two parts of a text they have the possibility to improve readers understanding of hypertextual documents. The success of such a technique however this may depend on the coherence of the reader's model of the text, and a discussion of the psychological factors that are believed to be involved is beyond the scope of this article.
References for works cited in this text chunk appear below. References for all works cited are available in a separate chunk.
Article No. 237, 2004-02-03, 2004.
Opinion presented as truth in alphabetical order.
& .