Database Reference
In-Depth Information
connect data to views. Design typically occurs directly within the interface that
contains data views, and often take effect immediately without the need for
a separate compilation or build stage. This live, amodal approach to interface
design allows users to switch rapidly between building and browsing tasks during
exploration and analysis. The result is a form of exploration that is free form and
open-ended, particularly during initial inspection of newly encountered data sets.
IVEE [2], DEVise [58], DataSplash [71], Snap-Together Visualization [69],
GeoVISTA Studio [88], Improvise [102], and Tableau/Show Me [60] are a few
of many well-known visualization environments that support open-ended data
exploration to various degrees. Such environments typically consist of a graphic
user interface on top of a library of visualization components which may or
may not be exposed as a visualization programming toolkit in its own right.
This combination of user interface and underlying library can enable open-ended
exploration in a very broad sense if it bridges the activities of visualization users
performing various roles with different levels of expertise, whether as individuals
or in collaborative groups.
To connect developers and designers, a key advantage for open-ended explo-
ration is an extensible library that provides an application programming inter-
face (API) for adding new software modules for various visualization components
(including data access, queries and other data transformation algorithms, views,
and visual data encodings). In particular, the most useful APIs support the def-
inition of new data transformation operators—including appropriate input and
output data object types—that give designers the ability to express rich relation-
ships between data, queries, and views. This requirement is essential for applying
newly discovered visualization techniques to emerging sources and forms of in-
formation, without needing to constantly architect and implement new toolkits
(and retrain visualization designers in their use).
To connect designers with users, the user interface must support the abil-
ity to access data sets (and metadata) from local or remote sources in various
formats, create and position views on the screen, specify how navigation and
selection affects views, specify queries on data, parameterize queries in terms of
interaction, and attach data sets and queries to views. In particular, designers
should be able to specify the appearance and behavior of their visualizations
directly within the user interface, without resorting to programming or other
workarounds for interface limitations. To do otherwise would effectively require
that designers be trained as developers.
User interfaces that truly support open-ended exploration would exceed the
requirements of basic visualization design and operation by: supporting live
building of complete browser interfaces, including immediate designing, debug-
ging, and testing of intended functionality; facilitating collaboration between
end-users and designers to turn analytical questions into structural changes
(through remote, nearby, or side-by-side efforts to communicate and effect rapid
visualization prototyping and polishing); and enabling rapid switching between
building and browsing to perform more extensive exploratory visualization by
modifying visualization views and queries on the fly. In particular, it is highly de-
sirable for explorers to be able to see all raw data quickly to make decisions about
Search WWH ::
Custom Search