Information Technology Reference
In-Depth Information
different context spaces. After query evaluation, the DB Proxy Box receives all
context data that includes the object
o
, that is, its
facets
of
o
. The result data
are sent to the Data Manager Box. The Data Manager Box then makes copies
of the template Box, the detailed structure of which is shown in the upper-
left of Figure 2. Each CSP Box materializes the context space by following the
beforementioned procedures. Users can choose the context space, enter it, and
follow the links depending on their interests.
3 Information Access for Chronicles
Vast amounts of records human experience are stored all over the world. To
understand and use such chronicles effectively, several relations embedded in
records must be found. This paper describes a chronicle analysis tool, which
provides several views of records from different viewpoints.
3.1 A Framework for Chronicle Analysis
A framework fr analyzing chronicle data consists of three layers. The first layer
provides a function for filtering data by specifying conditions. A parameter
scope
consists of two kinds of queries: a period and a viewpoint. A period is a query
concerning the date attribute, and a viewpoint is a query concerning the other
attributes.
The second layer visualizes the distribution of filtered data as a combination
of matrix and bar chart graphs. If users can find interesting patterns on the
distribution map in the second layer, the third layer represents transitions of
relations among focal attribute values by using the
topic sequence
technique [4].
The focal attribute is specified by a parameter focus. A
topic sequence
is regarded
as a graphical plot of a series of text chunks. Each chunk is visualized by a
word
colony
[3], which visualizes the dependency relationships among terms in a text
as a graph. A
word colony
is regarded as a visual abstract of a relation among
terms in a text. A
topic sequence
is formed by the connections of
word colonies
.
A
topic sequence
provides overviews of the topical change of text chunk content
over time.
At the beginning, users see the patterns of distribution for whole data. These
patterns give an overview of a target data set. If users can find interesting pat-
terns on the distribution map by selecting an attribute for the vertical axis, then
they can narrow their interest down by filtering data or they can see more de-
tailed information by visualizing a transition of interesting attribute values. The
distribution map or transitions of attribute values suggest users' next keywords
for specifying another viewpoint. A reiteration of this process corresponds to
a story/hypothesis/interpretation generation cycle. If users can find interesting
patterns of transitions of attribute values, then these patterns form the founda-
tions of new stories.
