Contexts, Narratives, and Interactive Visual Analysis of Names in the Japanese Hyohanki Diary

Visualization tools for the discipline of history
are being used increasingly. Historians can
explore their efficacy as both educational
instruments and platforms for displaying
research findings. In order to obtain patterns
of significance, historian’s work involves
manipulating, memorizing, and analyzing
substantial quantities of information from the
series of documents at their disposal. In
so doing, a central purpose emerges: the
construction of a narrative that best fits into the
representation of the past.
Visualization tools have proved to be effective
for facilitating users’ analytical tasks (Heer et al.,
2008; Heer et al., 2009; Weizhong and Chen,
2007, James and Cook, 2005). Moreover, they
have proved to be fruitful in the context of the
digital humanities (Bonnet, 2004; Dalen-Oskam
and Zundert, 2004). Recently, we proposed
a visualization system for analyzing aristocrat
names in a Japanese historical diary called
Hyohanki (Toledo et al., 2009). In our system,
the stacked graph is utilized to analyze the time
series of those names. Stacked graphs, stacking
time series on top of each other, are a useful
method to time series visualization, resulting
in a visual summation of time series values
that provides an aggregate view stratified by
individual series. Projects such as NameVoyager
(Wattenberg et al., 2005) and sense.us (Heer
et al., 2009) used animated stacked graphs to
explore demographic data.
In this paper, to preserve contexts through
the stacked graph usage, we propose an
extension of our previous work. Our system
provides two functionalities: an interaction
control for saving, querying, and deleting
views, as well as a dynamic repository of
views representing the context of the stacked
graph usage. Keeping useful contexts facilitates
coherent narratives (Bonnet, 2004). To support
historians’ efforts in keeping contexts through
interactive visual analysis of time series, both
functionalities aim at recording how users
receive units of information appropriated to
construct narratives.
To analyze the end-user perception of our
system, we conducted a heuristic evaluation
(Nielsen, 1992) with a domain expert who
explored the data using the tool. This heuristic
evaluation was a form of user study in which
the expert reviewed the system to suggest
advantages and disadvantages against the new
functionalities. This approach helps to further
elucidate the requirements and how the system
meets experts’ needs. The results provide useful
guidance for highlighting known historical facts
as well as hints to unknown historical facts.
1. Methodology
Our visualization system is based on a time
series set containing the quantitative analysis
of name occurrences in the Hyohanki diary.
The set contains 121 time series in a timescale
spanning from 1132 to 1171. Some parts of
the diary suffer from missing data; for that
reason, that period includes only the years 1132,
1139, 1149, 1152-1158 and 1166-1171. Likewise,
the data has been normalized in order to
measure the percentage relative value of the
number of occurrences of a given name, in
a given year, with the total percentage of
that year. Additionally, using the Euclidean
distance metric, we calculated the trends'
similarity between time series. For each name,
we recorded their five most similar trends.
The method used to visualize the data is
straightforward. Given a set of aristocrat names’
time series, a stacked graph is produced (Fig.
1a). The x axis corresponds to year and the y
axis to the occurrence ratio — in percentage, as