Introduction: Encoding System Design in the Context of User-Centered
Design
Writing about the Cypress project at the University of California at
Berkeley, Nancy Van House et al. state that "digital libraries can be
described and evaluated on three key components: contents, functionality,
and interface"°. The writers lament the fact that of these three, contents
and functionality are not commonly enough evaluated. For a digital resource
whose function depends to a high degree on the encoding of its contents, we
can extend these observations and suggest that user-centered design and
evaluation of functionality must begin with the design of the encoding
system itself. This is increasingly true as the encoding system becomes more
complex. For a project like the Brown University Women Writers Project
(WWP), whose research focuses on highly detailed encoding, our central
problems are how to provide encoding which will support anticipated user
needs; how to find a delivery system which can use this encoding; and how to
design an interface which makes this encoding useful to the user. Since the
topic of user interface design deserves its own paper, and since there are
already many studies of that subject°, this paper focuses on WWP encoding decisions rather than
on the choice and design of the delivery system°.
Moreover, the aforementioned studies are generalized, rather than focused on
humanities textbase design. Studies that do focus on humanities projects are
scarce, and the ones that do exist°
do not specifically address the issue of how the kind of markup offered in
the Text Encoding Initiative (TEI) Guidelines might best be utilized in
encoding and delivery. We are interested in exploring what components(s) of
TEI markup are actually necessary to fulfill scholarly user needs. For this
reason (and because it is not clear how much one could generalize from
others' studies) the WWP has performed our own assessment of users' needs
via beta-testing (prior to the textbase release date of August 1999) and via
a Mellon Foundation-funded user survey.
Defining "Rich" Text Encoding
Within the context of the TEI, the definition of "rich" encoding occurs on a
spectrum. The "lightest" involves transcribing from a modern standard
edition of a work using minimal phrase-level elements, ignoring original
pagination, lineation, and typography. The other extreme is to transcribe
several different versions of a work, marking them up so as to enable
collation/comparison, providing detailed analysis of handwritten
additions/deletions, and providing markup to enable linguistic and/or
metrical analysis. The WWP falls in the middle of these two extremes. We
provide robust phrase-level markup, preserve original errors, typography,
abbreviations, spelling, lineation, pagination, forme work information, and
basic information about handwritten additions/deletions. We also record a
few aspects of the original rendition (e.g., case, slant, and alignment, but
not kerning, leading, or type size).
Arguments For and Against Rich TEI Markup
One of the primary arguments against rich text encoding is that it exceeds
actual scholarly requirements, providing unnecessary detail at prohibitive
cost. Part of the problem with this argument is that scholarly requirements
vary considerably from discipline to discipline; historians may chiefly need
simple search and retrieval tools, while literary scholars may need features
which support more detailed analysis of the text. In addition, a diversity
of potential uses may also require more detailed encoding to accommodate
different needs. Thus, while scholars engaged in historical or literary
research value a transcription which reproduces the original lineation,
pagination and spelling, teachers may need a version with some degree of
spelling modernization or regularized typography for their students. The
WWP's user survey (http://www.wwp.brown.edu/project/rwo/Evaluation.html)
suggests strongly that users expect the textbase to function both for
research and for teaching, and that they do require a transcription which
reproduces the basic material details of the text as well as its structure
and content.
Justifying the Cost of Rich Text Encoding
Very few projects provide rich, detailed markup of primary source texts
because the expense of this level of markup is regarded as an obstacle to
getting important materials online, and greater priority is given to the
quantity of material encoded than to the detail of the encoding. While the
WWP also feels these pressures, we feel that the actual added cost and
effort are still matters for research. Our working hypothesis is that rich
markup is necessary in order to provide a research environment within which
users feel empowered and at home. Rather than thinking of rich text markup
as adding levels of arcane complexity which will never be used, we feel that
its level of detail anticipates the users' basic expectations about how text
should be represented and manipulated. Therefore, from our point of view the
question is not how to justify the costs associated with rich text encoding
but rather (given that rich text encoding is a necessity) what might be the
best way to minimize those costs. These methods of cost minimization will be
addressed at length in the final version of this paper.
Conclusions
While not decisive, our user survey and beta-testing results so far confirm
our hypothesis that rich text markup is essential for providing a textbase
which is a genuinely useful teaching and research tool. Moreover, these
results indicate the direction future testing should take. Future evaluation
and design iterations must focus even more on user-centered testing and
evaluation. In particular, as Seaman suggests°, pro-actively
educating users will help greatly in solving many still-unanswered
questions.