Information Cartography: A
proposed model for access to heterogeneous end-user databases. ASIS&T
SIG/CR Idea Mart, November 12, 2000, Chicago, IL
Information Cartography:
School of Information Studies
Syracuse University
4-206 Center
for Science and Technology
Syracuse, NY 13244-4100
phone: 315-469-2858
fax: 315-443-5806
e-mail: swpaling@mailbox.syr.edu
Maps are among our best information systems. They
require little documentation and are commonly used and understood. In contrast,
many systems of classification seem to lack this acceptance and ease of use.
Anecdotal evidence suggests that this is particularly true of the way
collections of databases are classified for online browsing on library Web
sites. This paper argues that some of the characteristics that make maps easily
usable can be applied to collections of databases. Those characteristics
include logical grouping of information, the ability to move smoothly between
levels of data, and consistent amounts of data at different levels of
representation.
This paper will take a conceptual approach in
discussing factors that make maps easy-to-use and readily acceptable, and will
sketch some of the implications those factors might have for the classification
of online databases. It will start with a description of several cartographic
phenomena and their utility. After that it will discuss how these phenomena
might be applied to information systems, and will finish with a discussion of
what such a cartographically inspired classification system might look like and
how it might be built.
Maps create well-formed expectations. We are
familiar enough with terms like "road map" and "floor plan"
to know what their contents are even before we see them. Bibliographic
descriptions carry similar expectations. Anyone familiar with libraries can
quickly decipher a bibliographic description for an article, book, etc. Our
classificatory descriptions of databases, though, do not achieve this. They
tend to feature free-text descriptions with neither consistent elements nor
formatting. The elements that users are likely to refer to in distinguishing
between databases (topic, scope, features, etc.) are generally described in
non-standard language. This kind of classificatory standardization could help
reproduce some of the well-formed expectations that maps and bibliographic
descriptions engender.
Maps feature pan and zoom, two traits that
contribute to their usability. Pan represents the ability to scan across the
surface of a map (paper or electronic) to see what features lie next to each
other. Zoom allows the user to view selected parts of the map in greater
detail. These traits are reproduced inconsistently in collections of online
databases. When a user searches in a database, or views its description, the
system often does not give the user an effective way to pan to a logically
adjacent database or description. Collections of databases do provide some zoom
capabilities by allowing the user to move from a list of database descriptions
to a single database or small group, and then zoom in further to particular
documents. Looking at the way maps (particularly electronic maps) provide this
functionality, though, points up both problems and possible solutions in the
way that we classify databases, and with what those classifications afford to
users.
"Constant information density" is
another key trait of maps that can help elucidate the organization of online
databases. As a user zooms from one level of a map to another, they should see
a relatively constant amount of data. For example, as they zoom in on a city
map, the streets may become less densely packed on the map, but more of the
street names and other features will appear. Database descriptions, though,
typically demonstrate a relatively poor level of detail compared to the rich
descriptions of documents occurring within the databases. A user who has not
found what they need in one database sees a precipitous drop in the level of
detail when they zoom back out to a list of database descriptions.
The paper proposes Frames of Reference (FoR), a
hypothetical system that would address the issues raised here. FoR would use
frames in conventional Web browsers to help users maintain the original context
of their search. An important part of this context involves the way databases
are classified to reveal their relationships with each other. We know from both
research and anecdotal evidence that user needs often center around orientation
within a search process rather than in specific elements of the search itself.
FoR proposes keeping links to logically adjacent, i.e., similarly classified,
databases available to the user at all times. This would be similar to seeing
the rest of a map on the periphery even while concentrating on a specific part
of it.
The final part of the paper discusses how
emerging tools like the Extensible Markup Language (XML) can be used to
implement systems like FoR that could reproduce pan, zoom, and constant
information density in our classification of online databases. In contrast to HTML, which is a page-layout
language, XML allows semantic information to be embedded into Web pages. Such
information would allow the addition of standard classificatory descriptions of
databases. Logical groupings of databases could be built around these
descriptions, and even generated dynamically to suit the context of a search.
Questions: