Visualizing Hierarchy Changes by Dynamic Indented Plots

Michael Burch, Tanja Blascheck, Christine Louka and Daniel Weiskopf

VISUS, University of Stuttgart, Stuttgart, Germany

Keywords:

Dynamic Hierarchies, Indented Plots.

Abstract:

Visualizing static hierarchical organizations has been a focus of information visualization for many years,

but it still remains challenging to produce visual representations of evolving, i.e. dynamic hierarchies. In

this paper we extend the concept of indented plots to also support the depiction of changing hierarchies.

We exploit the concept of static diagrams in order to support the preservation of a viewer’s mental map.

Changes between subsequent hierarchies are precomputed and visually indicated by color coded straight or

curved lines depending on the type of change. Interactive features can be used to aggregate or collapse the

dynamic hierarchy data in the supported dimensions, i.e. the vertex as well as the time dimension on different

levels of hierarchical and temporal granularity. The usefulness of our technique is illustrated by means of

a bibliography dataset where we show the changes in the yearly preﬁx tree acquired by extracting words

occuring most frequently in more than 2,000,000 paper titles from the digital library DBLP.

1 INTRODUCTION

There are many application examples where hierar-

chical data comes into play. Software developers typi-

cally organize their system into packages, directories,

subdirectories, ﬁles, classes, functions/methods, and

code blocks on the ﬁnest level of hierarchical granu-

larity. Biologists structure all species living on earth

into a hierarchical structure which is updated from

time to time with newly classiﬁed organisms. As a

third example occurring words in paper titles can be

hierarchically organized into a preﬁx tree structure.

All these hierarchies are typically changing over

time due to software development, the evolution of

life, and progress in research articles. Since such a dy-

namic preﬁx tree structure contains frequent and also

abrupt changes it will be used as illustrative applica-

tion scenario in our work as a stress test.

In this paper we introduce a novel approach to vi-

sually depict the evolving behavior of hierarchical or-

ganizations. To allow good comparisons we use the

concept of time-to-space mapping, i.e. a static dis-

play showing the sequence of hierarchies in a side-

by-side representation according to the small mul-

tiples approach (Tufte, 1983). This stands in con-

trast to animation that only shows one hierarchy at a

time and smoothly transforms one into the other (Card

et al., 2006). It is also different from other approaches

surveyed by Graham and Kennedy (2010) such as

coloring (Munzner et al., 2003), matrix (van Ham,

2003), agglomeration (Burch and Diehl, 2006), 3D,

or atomic representations. When using animation, a

good layout algorithm has to be applied that takes

the aesthetic appearance of all hierarchies into ac-

count and additionally targets a good dynamic stabil-

ity of the animated sequence focusing on mental map

preservation (Misue et al., 1995) and a reduction of

cognitive load. Another drawback of animation is the

fact that the viewers have to rely on their limited short

term memory in order to make comparisons between

subsequent hierarchies. Furthermore, it is very dif-

ﬁcult to apply interactive features to an animated di-

agram. Similar perceptual and interaction drawbacks

hold for other representation styles as discussed in the

survey by Graham and Kennedy (2010).

In this work, instead, we make use of a sequence

of indented plots (Burch et al., 2010) placed next

to each other in the same orientation. Such plots

are redundant-free, compact, and scalable indented

curves depicting hierarchical structures by the con-

cept of indentation similar to the concept of pretty

printing applied by programmers. Differences be-

tween subsequent hierarchies are precomputed and

visually indicated by color coded straight or curved

links supporting viewers with mental map preserva-

tion by guiding them through the hierarchy changes

via explicit links and not by smooth animation. This

edge drawing approach has the beneﬁt that analysts

Burch M., Blascheck T., Louka C. and Weiskopf D..

Visualizing Hierarchy Changes by Dynamic Indented Plots.

DOI: 10.5220/0004652400910098

In Proceedings of the 5th International Conference on Information Visualization Theory and Applications (IVAPP-2014), pages 91-98

ISBN: 978-989-758-005-5

 2014 SCITEPRESS (Science and Technology Publications, Lda.)

can make comparisons visually, and not mentally, by

exploiting the capabilities and strengths of their visual

system and fast pattern recognition. As another ben-

eﬁt, interactive features are easy to apply because the

display is static.

Our visualization tool is designed in a way follow-

ing the Visual Information Seeking Mantra (Shneider-

man, 1996), i.e. ﬁrst we give an overview of all the

changes, then we allow zoom and ﬁlter operations,

and ﬁnally details on demand. We allow data aggre-

gation in the vertex as well as the time dimensions

and support the analyst by further interactive features

such as ﬁnding trails of hierarchy elements.

We illustrate the usefulness of our technique by

a case study exploring bibliography data from the

digital library DBLP (Ley, 2009): we ﬁrst extract

all stored paper titles, count the occurring words in

each year, generate a preﬁx tree of the most frequent

words in the Newick format for each year, compute

the changes of subsequent preﬁx trees, and ﬁnally vi-

sualize these results by our visualization tool.

2 RELATED WORK

Visualizing static hierarchical data has been a focus

of research for many years where visual and algorith-

mic scalability (McGufﬁn and Robert, 2010) as well

as readability and aesthetics have been taken into ac-

count as competing criteria to trade-off. Several vi-

sual metaphors exist to depict hierarchical organiza-

tions: Node-link diagrams clearly show the hierarchi-

cal structure by mapping objects of a hierarchy typi-

cally to circular shapes and by indicating parent-child

relationships as straight or orthogonal links connect-

ing the hierarchy elements (Eades, 1992; Reingold

and Tilford, 1981). A top-down layout has been eval-

uated as the best performing one for the task of lo-

cating the least common ancestor of a list of marked

leaf nodes, as found out by an eye tracking study con-

ducted by Burch et al. (2011a).

Treemaps apply the concept of nesting enclo-

sure to illustrate a hierarchy in a space-ﬁlling man-

ner (Johnson and Shneiderman, 1991; Shneiderman,

1992). Many variations exist such as squariﬁed (Bruls

et al., 2000), cushion (van Wijk and van de Wetering,

1999), or Voronoi Treemaps (Balzer et al., 2005; No-

caj and Brandes, 2012) to enhance the layout and to

make the hierarchy representations more readable and

understandable. Though treemaps use the given dis-

play space efﬁciently, it is difﬁcult to visually explore

the hierarchical structure in a better way than in node-

link diagrams and to use them for dynamic hierarchy

visualizations (Tu and Shen, 2007).

The metaphor of layered icicles (Kruskal and

Landwehr, 1983) applies the concept of stacking to

build a visual depiction of a hierarchical organization

which allows a small multiples visualization and edge

drawing approach but on the negative side inner ver-

tices are represented as bars being as long as the sum

of the bars of all the child nodes’ bars. A small multi-

ples icicle representation would hence produce more

crossings and overlaps.

All these approaches have one problem in com-

mon. There is no single representative element for

each vertex — inner vertex as well as leaf vertex

— on a one-dimensional line. Indented Pixel Tree

Plots (Burch et al., 2010) instead use indentation om-

nipresent in graphical ﬁle browsers or pretty printing

of source code. These plots allow the mapping of all

vertices to a unique representative one-dimensional

position and can hence be laid out in a side-by-side

manner allowing good comparisons between subse-

quent hierarchies. As another beneﬁt, these plots al-

low the attachment of additional information to each

vertex in an aligned way (Burch et al., 2012) and scale

to very huge and deep hierarchies still sufﬁciently rep-

resenting the hierarchical structure. This mapping al-

lows us to enrich the diagram by color coded links

showing the differences and changes over time with

a reduced overlap of links with the sequence of hier-

archy visualizations — denoted as edge drawing ap-

proach in the survey by Graham and Kennedy (2010).

Generally, in graph visualization there also exist

two major concepts to represent the dynamic behavior

of a graph. Either animation can be used that shows

each timestep in a single frame and then smoothly

animates this toward the subsequent one by keeping

a high degree of dynamic stability (Diehl and G

org,

2002; Frishman and Tal, 2007). This requires quite

complex layout algorithms that follow aesthetic graph

drawing criteria (Purchase, 1997) for each static graph

in the sequence. Orthogonally to that, aesthetic cri-

teria for the dynamics (Beck et al., 2009) have to

be followed in order to preserve a viewer’s men-

tal map. This concept is referred to as time-to-time

mapping whereas a time-to-space mapping stands for

a representation of the inherent time dimension to

space (Burch and Diehl, 2008; Burch et al., 2011b).

For dynamic hierarchy visualization we could fol-

low a similar concept: We use a time-to-space map-

ping by exploiting the visual metaphor of indentation

by vertically aligning the hierarchies on parallel axes.

To achieve a compact and connected layout we do not

leave gaps for the alignment but use connected in-

dented plots and show the changes over time in the

dynamic hierarchies by colored straight and curved

lines similar to the concept used in CodeFlows (Telea

IVAPP2014-InternationalConferenceonInformationVisualizationTheoryandApplications

Figure 1: Hierarchy of depth 2 consisting of 10 vertices: 1

designated root vertex F, 2 inner vertices A and E, and 7 leaf

vertices M,N,P,B,C,D, and Q.

(a) (b)

Figure 2: Node changes: (a) Deletion of node D from orig-

inal hierarchy (Figure 1). (b) Addition of node O. (c) Ex-

change of positions of sibling nodes M and P. (d) Modiﬁed

position of node M.

and Auber, 2008). This approach is most related to

our technique but it is restricted to showing source

code (indented metaphor) and not hierarchies in gen-

eral. Moreover, the code representation is mirrored at

the vertical axes and the depth is not indicated by pure

indentation but differently long bars for each line of

code. Also collapsing and expanding features in the

time and vertex dimension are not supported. This

mapping allows juxtaposition of subsequent hierar-

chies as also applied by (Holten and van Wijk, 2008)

and (Munzner et al., 2003). However, in those papers

only the changes among leaf vertices are displayable.

In contrast, our approach is extended to show changes

among any kind of vertex — inner as well as leaf ver-

tices.

3 DATA MODEL

We model a hierarchy as an acyclic directed graph

H = (V, E)

where V denotes the set of vertices and E ⊂ V × V

the set of directed edges, i.e. parent-child relation-

ships. One vertex is the designated root vertex and

the complete hierarchy is always directed from this

root vertex to the leaf vertices.

A dynamic hierarchy H of length n ∈ N is modeled

as a sequence of subsequent hierarchies

H := H

, . . . , H

where H

= (V

, E

), V

is the i-th vertex set, and

⊂ V

×V

is the i-th edge set, i.e. the parent-child

relations of the i-th hierarchy in the sequence.

3.1 Newick Format

Our visualization tool uses the Newick data format for

storing hierarchy data and for representing hierarchies

in computer-interpretable form. The correspondence

between a hierarchy and the nesting of parentheses

is exploited in this concept. A sequence of hierar-

chies is given by a set of ﬁles containing one hierarchy

each. Our Java-based visualization tool ﬁrst extracts

all hierarchies from the given ﬁles and stores them

in the corresponding internal data structures. The

Newick format can easily be extended to allow addi-

tional information attached to the vertices and also to

the edges such as distance functions as commonly ex-

isting in the ﬁeld of bioinformatics for modeling phy-

logenetic trees. The tool is able to handle additional

data attachments by just introducing the correspond-

ing parsing interfaces.

3.2 Hierarchy Comparison

For the hierarchy comparison two successive H

and

i+1

of H are ﬁrst transformed by depth-ﬁrst traversal

into two vertex lists L

and L

i+1

. Since all hierarchy

elements are given a unique identiﬁer a simple algo-

rithm can be used to compute changed nodes between

two compared hierarchies. Both representative lists

are processed and three new lists are generated for

added, removed, exchanged, and modiﬁed hierarchy

elements (see Figure 1 and Figures 2 (a) - (d)). The

contextual information is important for this visualiza-

tion, which means that we additionally store where

in the list vertices have been added or deleted. Fur-

thermore, for a changed position we need to store the

position of the vertex in list L

and also the new po-

sition in list L

i+1

. If there are several vertices with

the same textual description we additionally take the

depth and contextual information into account for a

good comparison.

4 VISUALIZATION TECHNIQUE

The approach of mapping static hierarchical data to

one-dimensional indented curves is extended to visu-

ally encode time-varying hierarchies in a static dia-

gram giving a good overview of both dimensions in

the data: vertices and time. Parent-child relationships

are visually expressed by the idea of indentation. In

VisualizingHierarchyChangesbyDynamicIndentedPlots

(a) (b)

Figure 3: A hierarchical dataset presented in two visual

metaphors: (a) Node-link diagram in top-down traditional

layout. (b) Indented plot using color coding for visually en-

coding depth information.

this section we ﬁrst introduce the general idea of in-

dented plots, show how this approach can be extended

to the dynamic case, explain how changes and differ-

ences are visualized, and ﬁnally propose several in-

teractive features to manipulate and navigate through

the data.

4.1 Static and Dynamic Indented Plots

Since our visualization technique allows changes in

all nodes — inner nodes as well as leaf nodes — we

apply the concept of indented outlines that have a rep-

resentative hierarchy element at a single 1D line. Fol-

lowing this visual metaphor we are able to display a

sequence of hierarchies in a scalable side-by-side rep-

resentation and plot occurring links in between.

Figure 3 illustrates how a traditional node-link di-

agram looks like when an indented visual metaphor is

used. The color coding in Figure 3 (b) is applied to

better indicate the depth metric in the indented plot.

Here, a red to green color gradient is used, i.e. the root

node is represented in red color whereas the nodes lo-

cated deepest in the hierarchy are colored in green.

This design is helpful when only the indented plot

sequence is inspected and gives a better impression

about depth in the hierarchy. If edges are drawn this

feature can be switched off to not lead to irritations

when several color codings are used on top of each

other. Instead, a black-to-white color or transparency

gradient can be used instead.

For the time-varying hierarchy data we use a time-

to-space mapping by encoding each hierarchy on one

indented plot in a side-by-side view to parallel verti-

cal lines similar to the concept used in parallel coordi-

nates plots (Inselberg and Dimsdale, 1990); see Fig-

ure 4 for an illustration of such a side-by-side view.

Figure 4: Dynamic hierarchy visualization of an evolving

preﬁx tree structure generated from frequently occurring

words in more than 2,000,000 paper titles extracted from

DBLP. Only vertices matching a speciﬁc text pattern are

taken into account.

4.2 Change Phenomena in Hierarchies

There are several change phenomena that may occur

when comparing two different hierarchies. In our de-

sign we decided to use red for removed objects, green

for added objects, and blue for persisting but moved

objects.

• Deletion of a node: If a node disappears from a

hierarchy H

to a subsequent hierarchy H

i+1

indicate this by two red straight lines starting at

the position in hierarchy H

where the node is still

present and pointing to the position in hierarchy

i+1

where this speciﬁc node was originally lo-

cated, see change from Figure 1 to Figure 2 (a)

and Figure 4.

• Addition of a node: If a node is added in a hierar-

chy H

i+1

that was not present in hierarchy H

draw two green straight lines that originate at the

position of this added node in hierarchy H

and

point to the position where it is located now in

hierarchy H

i+1

, see change from Figure 2 (a) to

Figure 2 (b) and Figure 4.

• Exchange of node positions: The exchange of

two positions can be modeled by two modiﬁed

node positions and hence, two twisted blue col-

ored curved lines are drawn between hierarchy H

and hierarchy H

i+1

indicating the exchange, see

change from Figure 2 (b) to Figure 2 (c) and Fig-

ure 4.

• Modiﬁed node position: If a node changes its

position in a displayed hierarchy H

to a subse-

quent hierarchy H

i+1

this is indicated by curved

blue colored lines connecting both representative

objects of the same node, see change from Fig-

ure 2 (c) to Figure 2 (d) and Figure 4.

IVAPP2014-InternationalConferenceonInformationVisualizationTheoryandApplications

All other more complex hierarchy changes can be

achieved by combining these simple rules.

Figure 4 shows how the change phenomena illus-

trated in Figures 2 (a) - (d) can be represented by a

side-by-side view of indented plots with color coded

straight or curved links. The color coded links — red,

green, and blue ones — can serve as an overview to

get a ﬁrst impression if and in what time step more

vertices are added, deleted, exchanged, or modiﬁed.

Our tool also allows the user to obtain some statis-

tics about these change phenomena indicated as color

coded bar charts, see Figure 5 where the graphical

user interface (GUI) of our visualization tool is illus-

trated.

4.3 Interactive Features

The tool supports several interactive features to ex-

plore the dynamic hierarchical data such as collaps-

ing and expanding speciﬁc nodes, getting information

on a speciﬁc element, and selecting a subregion for

a hierarchy on a larger scale for a better exploration.

Some of the tool’s features and functionalities are ex-

plained in the following:

• Region selection: Parts of the indented plot can

be selected and this region is then highlighted and

displayed to the right hand side of the original hi-

erarchy in a larger scale.

• Hierarchy expanding/collapsing: Clicking on a

node allows the user to collapse or expand the

corresponding subhierarchy. When a node is col-

lapsed all its children are hidden, and whenever

the node is expanded again the children reappear.

• Text pattern search: Typing in certain text frag-

ments starts a search for the speciﬁed text as a

substring in the textual descriptions of the vertices

and highlights the corresponding nodes found, see

also Figure 4. This features can be used to ﬁnd

word trails for example.

• Zooming: A zoom in/out horizontal slider bar can

be used for a larger/smaller scale.

• Details-on-demand: By moving the mouse on

any element on the display screen a more de-

tailed information for a speciﬁc node like the node

name, number of children of the node, the parent

name, and the depth is displayed.

• Color coding: For a better visualization of the

different hierarchy levels, each level of nodes is

represented in a different color to distinguish it

from other elements. The used color gradients can

be changed at user’s demand.

• Change ﬁlter: Only the changes of interest can

be displayed, i.e. either added, deleted, modi-

ﬁed, or exchanged elements. This means that only

the colored links of interest are displayed to the

viewer.

• Delta percentage: The percentage of change for

added, deleted and modiﬁed/exchanged nodes be-

tween two consecutive hierarchies selected can be

displayed.

5 CASE STUDY

The DBLP bibliography dataset contains more than

2,000,000 paper titles to date. From this data, all

words occurring in the paper titles can be extracted

and preﬁx hierarchies can be generated for each year

starting in 1936 and ending up in 2012 resulting in 77

hierarchies.

Displaying the hierarchical organization of the

preﬁx trees as indented plots in a side-by-side rep-

resentation can give us interesting insights into

newly introduced, deleted, or moved words or word

groups in the lexicographically ordered and vertically

mapped tree structure. For this case study we ﬁrst

compute the 100 most frequently occurring words in

paper titles (stop words removed) for each year that

occur at least 10 times in this year. From these preﬁx

trees we compute the dynamic preﬁx tree differences.

By our visualization approach we are now able to see

both the hierarchical structure in each year as well as

the added, deleted, modiﬁed, or exchanged vertices.

This means that we are able to analyze how the pub-

lication behavior of researchers changes due to novel

ﬁndings and breaking hot topics or obsolete and unin-

teresting research areas.

Figure 5 shows the dynamic preﬁx tree structure

and the added, deleted, exchanged, and modiﬁed el-

ements for a selected time interval (years 1942 until

1951). The ﬁrst insight that we get is the fact that be-

fore the year 1948 there are not even 100 words that

occur at least 10 times in all of the paper titles pub-

lished. In 1944, there is a minimum of words occur-

ring at least 10 times. From 1942 to 1943 many words

are removed from the list of the top 100 and are not

replaced by new words occuring at least 10 times in

this year. We hypothesize that there was a hot topic

in these early years which is not researched anymore

that much in the following years. Having a look at our

visualization (Figure 5) we see the many red straight

lines starting at 1942 and ending at 1943 for example.

Using a ﬁlter function for showing only the deleted el-

ements we can now easily hover the mouse to the sin-

VisualizingHierarchyChangesbyDynamicIndentedPlots

Figure 5: The graphical user interface of our visualization tool for analyzing dynamic hierarchies. It is divided into four main

parts. The panel on the left hand side is responsible for parameter settings and ﬁlter values. The center panel shows the time-

varying hierarchies whereas the panel on the right hand side is used for data output such as details on demand and legends.

The bottom panel shows some statistical information about the vertices from different categories such as added, deleted, or

changed vertices.

gle lines and obtain textual details about the deleted

words:

For example, we see that the words ’types’,

’tarski’, ’paradox’, ’analysis’, and ’application’ do

not belong to the preﬁx tree of the top 100 words in

1943. On the other side, words like ’theorem’, ’the-

ory’, ’system’, ’set’, or ’sequence’ are still present in

1943 which can either be found out by inspecting the

blue colored curved lines or looking for empty gaps

which means nothing has changed there from one

year to the other.

We can also use the visualization for ﬁnding

longer word trails in the dynamic preﬁx trees. For this

we ﬁrst ﬁlter for modiﬁed or exchanged elements and

get only blue colored curved lines (added and deleted

ones are grayed out). By hovering the mouse to blue

colored lines or corresponding nodes we ﬁnd words

that occur for a longer time. The words ’calculus’ and

’calculi’ can be found which are present for several

subsequent years: ’calculus’ starts to occur at least 10

times in 1936 and is present until 1955, disappears

in 1956 and 1957, and reappears in 1958 and 1959

again. From 1960 on, it will never occur in the top

100 words. The plural form ’calculi’ ﬁrst occurs in

1939 and 1940 in the top 100, then from 1942 to 1945,

and again from 1947 until 1951. In later years (1956

until 1958) the word ’calculation’ as well as ’calcul’

occurs at least 10 times in the top 100. After 1960, all

of these words never occur again in the top 100 which

shows that research in this ﬁeld may have changed.

6 DISCUSSION

Our novel visualization approach is based on the vi-

sual metaphor of indentation used for representing hi-

erarchies in order to allow a redundant-free and com-

pact representation with a low data-ink ratio. Further-

more, we use the concept of time-to-space mapping

that has some beneﬁts when inspecting time-varying

phenomena in the hierarchical data such as trends,

countertrends, periodicities, temporal shifts, and also

anomalies. Beneﬁts and drawbacks of the proposed

visualization technique can be summarized in the fol-

lowing way.

For visualizing a static hierarchy several aspects

concerning beneﬁts and drawbacks can be discussed:

• Parent-child relationships: By using the con-

cept of indentation for representing a hierarchy

we do not need explicit links for illustrating

parent-child relationships as for example in node-

link diagrams. These are only shown implicitly

by indenting subhierarchies. This results in a

IVAPP2014-InternationalConferenceonInformationVisualizationTheoryandApplications

redundant-free diagram and additional crossings

and overlaps with the lines indicating changes be-

tween subsequent hierarchies are avoided. On

the negative side the hierarchical structure is not

that clearly visible as in traditional node-link dia-

grams.

• Representative elements: Indentation allows

representative elements for each hierarchy vertex

— inner as well as leaf vertices — mapped to a

single line. This mapping can be used for attach-

ing additional attributes or in our case showing ex-

plicit lines for changes between subsequent hier-

archies.

For dynamic hierarchy visualization several ben-

eﬁts and drawbacks can be found when compar-

ing time-to-time mappings (animation) with time-to-

space mappings (static display):

• Mental map preservation: We use a static rep-

resentation for the evolving hierarchies. This has

several beneﬁts over an animated diagram. When

abrupt changes occur in the sequence an animated

representation may lead to losing the mental map,

which may make the data unreadable. To solve

this problem in our static representation we use

a side-to-side hierarchy visualization and use ex-

plicit links to guide the eye as also described

by (Graham and Kennedy, 2010). This static rep-

resentation reduces cognitive load, allows for in-

teractive features, and reduces the runtime com-

plexity that would be needed in the animated ap-

proach to produce layouts beneﬁtting from high

dynamic stabilities.

• Time-dependent phenomena: Time-varying be-

haviors can be explored visually in a static repre-

sentation. In an animated sequence this seems to

be problematic since the viewers have to do the

comparisons by relying on their short term mem-

ory. Consequently, trends, countertrends, period-

icities, temporal shifts, or anomalies can be de-

tected much easier in a static representation of dy-

namic data.

• Application of interactions: An animated dia-

gram makes it difﬁcult to apply interactive fea-

tures. The animation has to be stopped ﬁrst and

then a graphical element can be interactively se-

lected and manipulated. This is easier to apply in

a static diagram where the viewer has much more

time to apply mouse interaction.

• Algorithmic complexity: In an animated dia-

gram for hierarchical data good layout strategies

have to be used in order to achieve dynamic stabil-

ity, a concept that is also used in graph animation.

This is required to preserve the viewer’s mental

map of the animated diagram. In a static plot

of dynamic data such complex algorithms are not

needed. Here, the vertex ordering is not changing

and hierarchies can be added on-the-ﬂy without

computing a totally new layout.

• Visual scalability: Animated diagrams only

show one hierarchy at a time and hence, beneﬁt

from visual scalability. The whole display can be

used to represent this hierarchy. In a side-by-side

visualization, i.e. a time-to-space mapping, dis-

play space is needed for many timesteps which

allows each single diagram to be displayed only

on a small region on screen.

7 CONCLUSIONS AND FUTURE

WORK

In this paper we introduced a visualization technique

for visualizing dynamic hierarchies in a static dia-

gram by making use of the indented outline metaphor.

Large hierarchies are represented by using the con-

cept of several indented plots in a side-by-side rep-

resentation. The changes and differences between

subsequent hierarchies are indicated by color coded

straight or curved lines. Interactive features can be

applied to further navigate in the data, to explore and

analyze it, and to aggregate it in the vertex and time

dimension. A case study for bibliography data from

the digital library DBLP was presented illustrating the

usefulness of the novel dynamic hierarchy visualiza-

tion.

In the future we plan to make the visualization

technique more scalable in the time dimension. More-

over, the vertex ordering for reducing cluttering of

the display by blue colored lines may be a good

strategy to be investigated. The continuation of the

lines across evolution is hard to follow in a cluttered

display and hence, occlusion problems may be re-

duced by computing a good hierarchy layout for each

timestep. Also, a user study should be conducted

investigating the performance of participants when

working with the novel technique. Exploring other

application examples for dynamic hierarchies are also

of interest for future work.

ACKNOWLEDGEMENTS

In part, this work was supported by Deutsche

Forschungsgemeinschaft (DFG) within SFB 716 /

D.5.

VisualizingHierarchyChangesbyDynamicIndentedPlots

REFERENCES

Balzer, M., Deussen, O., and Lewerentz, C. (2005). Voronoi

treemaps for the visualization of software metrics. In

Proceedings of Software Visualization, pages 165–172.

Beck, F., Burch, M., and Diehl, S. (2009). Towards an

aesthetic dimensions framework for dynamic graph visu-

alisations. In Proceedings of Information Visualisation,

pages 592–597.

Bruls, M., Huizing, K., and van Wijk, J. J. (2000). Squar-

iﬁed treemaps. In Proceedings of Joint Eurograph-

ics/IEEE TVCG Symposium on Visualization, pages 33–

42.

Burch, M. and Diehl, S. (2006). Trees in a Treemap: Vi-

sualizing multiple hierarchies. In Visualization and Data

Analysis, SPIE Press, 6060, pages 224–235.

Burch, M. and Diehl, S. (2008). TimeRadarTrees: Visual-

izing dynamic compound digraphs. Computer Graphics

Forum, 27(3):823–830.

Burch, M., Heinrich, J., Konevtsova, N., H

oferlin, M., and

Weiskopf, D. (2011a). Evaluation of traditional, or-

thogonal, and radial tree diagrams by an eye tracking

study. IEEE Transactions on Visualization and Computer

Graphics, 17(12):2440–2448.

Burch, M., Raschke, M., Greis, M., and Weiskopf, D.

(2012). Enriching indented pixel tree plots with node-

oriented quantitative, categorical, relational, and time-

series data. In Proceedings of Diagrams, pages 102–116.

Burch, M., Raschke, M., and Weiskopf, D. (2010). Indented

Pixel Tree Plots. In Proceedings of the International

Symposium on Visual Computing, pages 338–349.

Burch, M., Vehlow, C., Beck, F., Diehl, S., and Weiskopf,

D. (2011b). Parallel edge splatting for scalable dynamic

graph visualization. IEEE Transactions on Visualization

and Computer Graphics, 17(12):2344–2353.

Card, S. K., Suh, B., Pendleton, B. A., Heer, B., and Bod-

nar, J. W. (2006). Time tree: Exploring time changing

hierarchies. In IEEE VAST, pages 3–10.

Diehl, S. and G

org, C. (2002). Graphs, they are changing.

In Proceedings of Graph Drawing, pages 23–30.

Eades, P. (1992). Drawing free trees. Bulletin of the Institute

for Combinatorics and its Applications, 5(2):10–36.

Frishman, Y. and Tal, A. (2007). Online dynamic graph

drawing. In Proceedings of EuroVis, pages 75–82.

Graham, M. and Kennedy, J. B. (2010). A survey of

multiple tree visualisation. Information Visualization,

9(4):235–252.

Holten, D. and van Wijk, J. J. (2008). Visual comparison of

hierarchically organized data. Computer Graphics Fo-

rum, 27(3):759–766.

Inselberg, A. and Dimsdale, B. (1990). Parallel coordinates:

A tool for visualizing multi-dimensional geometry. In

IEEE Visualization, pages 361–378.

Johnson, B. and Shneiderman, B. (1991). Tree-Maps: A

space-ﬁlling approach to the visualization of hierarchical

information structures. In Proceedings of IEEE Visual-

ization Conference, pages 284–291.

Kruskal, J. and Landwehr, J. (1983). Icicle plots: Better

displays for hierarchical clustering. The American Statis-

tician, 37(2):162–168.

Ley, M. (2009). DBLP: Some lessons learned. Proceedings

of Very Large Data Bases, 2(2):1493–1500.

McGufﬁn, M. J. and Robert, J.-M. (2010). Quantifying the

space-efﬁciency of 2D graphical representations of trees.

Information Visualization, 9(2):115–140.

Misue, K., Eades, P., Lai, W., and Sugiyama, K. (1995).

Layout adjustment and the mental map. Journal of Visual

Languages and Computation, 6(2):183–210.

Munzner, T., Guimbretiere, F., Tasiran, S., Zhou, Y., and

Zhang, L. (2003). TreeJuxtaposer: Scalable tree com-

parison using focus+context with guaranteed visibility.

ACM Transactions on Graphics, 22:453–462.

Nocaj, A. and Brandes, U. (2012). Computing Voronoi

treemaps: Faster, simpler, and resolution-independent.

Computer Graphics Forum, 31(3):855–864.

Purchase, H. C. (1997). Which aesthetic has the greatest ef-

fect on human understanding? In Proceedings of Graph

Drawing, pages 248–261.

Reingold, E. M. and Tilford, J. S. (1981). Tidier drawings

of trees. IEEE Transactions on Software Engineering,

7(2):223–228.

Shneiderman, B. (1992). Tree visualization with tree-maps:

2-d space-ﬁlling approach. ACM Transactions on Graph-

ics, 11(1):92–99.

Shneiderman, B. (1996). The eyes have it: A task by data

type taxonomy for information visualizations. In IEEE

Symposium on Visual Languages, pages 336–343.

Telea, A. and Auber, D. (2008). Code Flows: Visualizing

structural evolution of source code. Computer Graphics

Forum, 27(3):831–838.

Tu, Y. and Shen, H.-W. (2007). Visualizing changes of hier-

archical data using treemaps. IEEE Transactions on Vi-

sualization and Computer Graphics, 13(6):1286–1293.

Tufte, E. R. (1983). The visual display of quantitative in-

formation. Cheshire, CO: Graphics Press, p. 197.

van Ham, F. (2003). Using multilevel call matrices in large

software projects. In INFOVIS, pages 227–232.

van Wijk, J. J. and van de Wetering, H. (1999). Cushion

treemaps: Visualization of hierarchical information. In

Proceedings of Information Visualization, pages 73–78.

IVAPP2014-InternationalConferenceonInformationVisualizationTheoryandApplications