A taxonomy of temporal data

visualization techniques
Chaouki Daassi, Laurence Nigay, Marie-Christine Fauvet

Laboratoire CLIPS-IMAG, Grenoble, France

{Chaouki.Daassi, Laurence.Nigay,
Marie-Christine.Fauvet}@imag.fr
Résumé
Plusieurs techniques de visualisation et d’interaction avec des
espaces de données temporelles ont été proposées dans la littérature.
Toutes ces techniques ont été conçues indépendamment les unes des
autres. Dans ce contexte, il est difficile d’explorer l’ensemble des
possibilités de conception de ce type de technique de visualisation.
Dans la perspective de résoudre ce problème, cet article présente une
taxonomie des techniques de visualisation de données temporelles.
Sur la base de l’étude de trente sept techniques et systèmes de
visualisation de données temporelles, notre taxonomie structure
l’espace de conception selon les étapes du processus de visualisation
de Chi.
Mots-clés : Techniques de Visualisation, Données Temporelles,
Processus de Visualisation, Taxonomie.

Abstract
A wide range of visualization techniques have been designed to
assist users to visually analyze and manipulate temporal data. All the
proposed techniques have been designed independently. In this
context it is therefore difficult to systematically explore the set of
possibilities as well as to thoroughly envision visualization
techniques of temporal data. Addressing this problem, we present a
taxonomy of visualization techniques of temporal data. Based on a
study of thirty-seven visual tools and visualization techniques of
temporal data, our taxonomy structures the design space according
to the Chi's visualization process steps.
Keywords: Visualization Techniques, Temporal Data, Visualization
Process, Taxonomy.
1. INTRODUCTION AND RELATED WORK
An object property is said to be temporal if its successive values are
meaningful and thus recorded. A value of a temporal property is a history
or a set temporal data, e.g. a function from a finite set of instants observed
at a fixed granularity (i.e. a time unit), to a set of values of a given type.
The domain and the range of a history are respectively called its temporal
and structural domain [11]. In the study reported in this paper, we focus
on histories whose structural values are of a quantitative type. Examples
of such history are the salary of an employee or the daily productions of a
firm. In the following, the expressions "temporal data" and "histories" are
considered to have the same meaning.
Many visualization techniques of temporal data have been proposed in
various application domains including the display of clinical data [26],
geographical data [20], hydrometric data [18], personal histories [23], and
for different purposes as for searching trends and patterns [12], exploring
program traces [25], analyzing log data [13], representing temporal
abstractions [27] or visualizing temporal association rules [24]. All the
proposed techniques have been designed independently. Within this
context, it is therefore difficult to compare the existing visualization
techniques and explore new designs. In this paper, we present a
classification space based on the steps of the Chi’s visualization process
to provide a systematic classification process of visualization techniques
of temporal data. The contribution of our classification space is two-fold:
• The ability of our framework to compare/classify existing visualization
techniques has great promise, especially in light of the rapid progress
that we are experiencing in the domain.
• By identifying and organizing the various aspects of visualization, our
framework should also help the designer to address the right design
questions and to envision future visualization techniques of temporal
data.
Few design and classification spaces have been proposed in the
literature. Silva et al. [29] classify the visualization techniques according
to two classes of features: visualization and interaction features.
Visualization features include the data characteristics (e.g. complex
entity, entity-relation, periodic pattern, pattern/trend) and the rendering
(snapshot view, focus+context, multiple calendars, user-defined display).
Interaction features are derived from Shneiderman’s taxonomy of tasks
[28] that is general and not dedicated to interaction with temporal data.
Complementary to this classification scheme, several time-specific
interaction tasks (e.g., merging two timelines in order to obtain a new
one, juxtaposing two timelines) for visualization techniques of temporal
data are proposed in [19]. Another taxonomy of temporal data
visualization techniques is described in [22]. The taxonomy identifies two
classes of visualization techniques: static versus dynamic. A technique is
called dynamic if the associated visual representation is time-dependent,
otherwise it is called static. In static techniques, the visual representation
of objects does not change, unless a modification is required by the user.
By contrast, in dynamic techniques the visual representation of objects
evolves according to the time of modification of their values.
Although these studies identify classes of visualization techniques,
they do not extend enough so as to be used as guidelines during the
design phase. Addressing this issue in the context of data visualization in
general, Chi [3] structures the visualization process into steps described in
section 2, from the data to their rendering. Techniques are then classified
according to the visualization process they implement: Doing so, Chi
shows the similarities and differences between them. However, the Chi's
study does not consider the specific role of time in visualizing temporal
data. In this paper we present a classification scheme of techniques
dedicated to histories (e.g. temporal data) visualization. The classification
criteria are based on the steps of the Chi’s visualization process. To
validate these criteria, we studied thirty-seven tools and techniques
dedicated to temporal data visualization [7]. However, in this paper only a
few of them are used as illustrations.
The rest of the paper is organized as follows. In Section 2, we describe
the Chi’s visualization process and in Section 3 we explain how we apply
this process to temporal data. Section 4 introduces our taxonomy of
temporal data visualization techniques. Finally, in section 5, we conclude
and sketch some further studies.

2. ED CHI’S VISUALIZATION PROCESS

As shown in Fig. 1, Chi [3] decomposes the visualization process into
three types of transformations manipulating four types of data. In this
paper, we adopt the point of view of Vernier [31] that structures the
visualization process of Chi into four steps namely: (1) data, (2) point of
view on the data, (3) visualization space and (4) point of view on the
visualization space.
Data. For example, as illustrated in Fig. 1, the data could be a set of
URLs.
 Point of view on the data. At this stage of the process, the data are
transformed onto an analytical representation. This transformation is
generally dependent on the data structures. Indeed the point of view on
the data is obtained by defining an analytical abstraction, in other words
by obtaining meta-information, information related to the data such as
their organization (hierarchical, sequential, etc.) and by filtering the initial
space. For example, in Fig. 1, the web pages are retrieved from the list of
URLs (initial data space), then an image is generated for each web page.
Visualization space. The visualization space is defined by a
transformation of visualization, which maps an analytical representation
onto values ready to be displayed. In the visualization process shown in
Fig. 1, the images (the point of view on the data) are ordered (an ordered
list of images). Moreover, this transformation fixes the characteristics of
the graphical space in which the data are visualized in terms of the
number of its dimensions.

F IG . 1 – Chi's visualization process.

Point of view on the visualization space. This step produces the

perceptible rendering. It results from a visual mapping transformation
which maps the directly displayable values onto a graphical
representation (a view). In the example of Fig. 1, a book is defined for
presenting the list of web pages. So, at the visualization space step, the
designer decides on the characteristics of the graphical space. According
to these characteristics, the data values of the previous step are mapped
onto graphical signs. Moreover, at this step, a deformation function of the
graphical space could be applied, for example to visualize a large amount
of data.
In addition to the decomposition of the visualization process into three
types of transformations manipulating four types of data, Chi defines
operators dedicated to each step. For instance, operators related to the
“data” step include data filtering, adding new sets of data and the Fourier
transformation of images. Operators for the “point of view on the
visualization space” step include rotation, translation, enlargement of
graphical objects and the positioning and orientation of a camera in a 3D
scene. Other operators are defined for the two steps “point of view on
data” and “visualization space”.
Having presented the Chi’s visualization process, we now apply this
process to visualize temporal data.

3. VISUALIZATION PROCESS OF TEMPORAL DATA

As explained in Section 1, histories or temporal data are defined by
two domains, i.e., the temporal and structural domains. Based on these
two domains, we define the visualization process of a temporal data space
as an association between two visualization processes: one is dedicated to
the temporal domain of the history to be visualized, while the other is
related to the structural domain of the history. The visualization process
of structural values depends on the type of these values. Studying this
process consists of studying the visualization process of data in general,
which is out of the scope of this paper. We therefore only focus on the
visualization process of the temporal domain. Fig. 2 shows how we apply
the Chi’s visualization process to the temporal domain of histories: four
steps are defined namely time, point of view on time, time space and
point of view on the time space.
Time. This step represents the time values to be visualized. In the
example of Fig. 2, we consider a discrete, linear and bounded time model
in which the time line is structured in a multi-granular way by means of
time units (Year, Month, Day, Hour, etc). Time values are then observed
at a particular granularity fixed by a time unit [32] (in the example of Fig.
2, the time unit is Day).

F IG . 2 – Visualization process of a temporal domain.

Point of view on time. This step is defined by a transformation of time

values. This transformation extracts from the time values the way time is
represented: (i) in which time unit are the time values observed? (ii) are
the time values expressed in a particular unit system? A unit system is a
sequence of comparable1 units ordered from the coarsest one to the finest
one, e.g. [Year, Month, Day]. In the example of Fig. 2, time is considered
as multi-granular and linear.
Time space. At this step, displayable space of the time values is
defined. For example, if time values are observed at unit Hour and
expressed in unit system [Day, Hour] (the analytical representation of
“the point of view on time” step), they could be mapped onto a two-
dimensional representation. One dimension represents the days (Day is a

1
One time unit U1 is said to be comparable to another time unit U2, only if each value of
U1 is composed of a fixed number of U2. The time unit U1 is said to be coarser than the
time unit U2, and U2 is finer than U1. A concrete example of regular units is Year (coarser
unit) and Month (finer unit). Each year contains twelve months.
unit coarser than Hour, and comparable with Hour), while the other
represents the hours. In the example of Fig. 2, each time unit is
considered as linear: one time axis is then associated with each of them.
Point of view on the time space. The point of view on the time space
refers to the rendering of the temporal domain of the history to be
visualized. At this step, the perceptible forms of time are implemented.
These forms must be studied according to the human perception of time.
For instance, as pointed out by Tufte [30, p. 186], human eyes are able to
detect deviation of the horizon, making easier the perception and
interpretation of the horizontal representations of temporal data.
The choice of a form is strongly influenced by the point of view on
time that has been adopted (Step 2 “Point of view on time”). Indeed, the
point of view on time identifies the characteristics of the temporal values
to be visualized. For example, a graphical representation of time in a
cyclic form as illustrated by the spiral shown in Fig. 3-a could be used
because the history is observed at a given unit system and could be
viewed as periodic: this characteristic was identified during the step
“Point of view on time”. In Fig. 3-a, the time line is shown as a spiral and
a temporal value corresponds to the intersection between the spiral and
one of the twelve radii (one for each month). Each structural value is
represented by a circle, whose size depends on the value it is associated
with (higher is the value, larger is the circle). Fig. 3-b (a simplified
representation of the TimeSlider technique [16]) illustrates another
representation of time where points of time are distributed irregularly on
an axis in order to render a logarithmic perception of time [5]. Indeed in
Fig. 3-b the time scale is linear at the central part of the timeline and is
based on a uniform time unit (e.g., a day in Fig. 3-b) whereas the time
scale is non-linear (time steps are exponential) at the edges. Finally a
linear representation is another classical way to represent time as depicted
in Fig. 2 where four timelines [10] are used to interact with the time
space: one timeline for each time unit.
Some visualization techniques rely on various forms of time used
simultaneously. For example the Spiral Calendar Visualizer [21] includes
two points of view on time: multi-granular and cyclic. The values of each
time unit are represented as a calendar (multi-granular representation of
time). Thus, years are first represented in a view, then one selected year is
displayed in another view and a month of this year is displayed in another
view and so on. All the views are represented in a spiral form (cyclic
perception of time). Using simultaneously these two forms of time
reinforces the user’s perception of the multi-granular structure of time
while highlighting cycles.
 -a- Spiral representation of time [2] -b- TimeSlider [16]
F IG . 3 – Time representations.

4. A TAXONOMY BASED ON THE STEPS OF THE

VISUALIZATION PROCESS
Our taxonomy of temporal data visualization techniques is based on
the different anchoring points between the visualization processes of the
temporal domain (as described in the previous section) and that of the
structural domain. Considering the different anchoring points between the
two visualization processes enables us to define seven classes of
visualization techniques, described in Fig. 4. These configurations range
between two extremes: the total fusion of the two processes (Fig. 4-a) and
the total splitting of the two processes (Fig. 4-g).
A point of contact (i.e., the first common step) between the two
visualization processes is defined according to the operators that are
applied at the corresponding step. These operators, as defined by Chi, are
specific to each of the four visualization process steps. For example
translation or rotation are operators dedicated to the step “point of view
on the visualization space”. If at a given step, one operator has an impact
on the data manipulated by both processes then we consider the
corresponding step as a point of contact. One operator is a condition
sufficient to declare the corresponding step point of contact. It is worth
noting that some operators may remain specific to only one of the two
processes. Once a step is defined as a point of contact, all the following
steps in both processes are then merged. Indeed, to the best of our
knowledge there does not exist any approach corresponding to the
configuration of Fig. 5.

F IG . 4 – Five classes of temporal data visualization techniques according to

their visualization process.

Total fusion or fusion at the step “point of view on the data”. Some
visualization techniques of temporal data result from a total fusion of both
visualization processes. We distinguish two cases respectively shown in
Fig. 4-a and Fig. 4-b. The configuration depicted in Fig. 4-a is often used
to visualize data whose temporal domain is implicit: the representation of
time is then implicit. Examples of such histories include videos.
Conversely, the configuration shown in Fig. 4-b is used to visualize
histories whose temporal domain is explicit. Time is then explicit,
enabling the dissociation of the two history domains (i.e. the temporal and
structural domains). Visualization techniques could then be classified
depending on whether the time representation is implicit or explicit.
F IG . 5 – The separation of the two visualization processes after a crossing-
point: No existing visualization technique corresponds to this configuration.

Considering the visualization techniques whose representation of the

temporal domain is implicit, we distinguish two categories of techniques,
dynamic vs. static [22] as defined in Section 1: (i) dynamic techniques
whose visual representation changes automatically. Campos et al. [1]
propose an animation model to support exploratory analysis of dynamic
environments. Animation is frequently used to visualize spatio-temporal
data as with the TimeMap data viewer [14]; (ii) static techniques whose
visual representation does not change automatically over time. The
information mural technique [15] (Fig. 6) is a static representation of
temporal data that belongs to this latter class.

F IG . 6 – Information Mural [15], from

http://www.cc.gatech.edu/gvu/softviz/infoviz/information_mural.html.
 The information mural technique supports visualizations of large
amounts of data: the number of data values is greater than the number of
pixels available on screen. This technique transforms the initial data space
which is defined with M*N pixels into a smaller space of I*J pixels while
preserving a similar global view of the initial data space. Based on this
technique, Fig. 6 presents a visualization of the number of daily recorded
sun spots observed between 1850 and 1993 which is over 52,000
readings. As the temporal domain is not explicitly represented, both the
“time space” and the “point of view on the time space” steps of the
temporal domain visualization process are not considered.
The diagram representation of Fig. 7 is yet another static technique, but
with an explicit representation of the temporal domain. It is commonly
used to visualize quantitative data, temporal intervals and temporal
relations [4] [17]. In these representations, time is usually represented on
the horizontal axis and structural values on the vertical axis. In the
example of Fig. 7, the technique visualizes the air pollution measures
over time of NO (nitric oxide) and NO2 (nitrogen dioxide). Structural
values are represented as a curve. Fig. 8 represents its corresponding
visualization process (configuration –b- in Fig. 4): the visualization
process of the temporal domain and that of the structural domain are
combined at the “point of view on the data” step. Consequently, the
representation of the structural domain (the curve) and that of the
temporal domain (the time axis) belong to the same graphical space: both
domains are visualized within a two-dimensional graphical space. In the
visualization process of the temporal domain of Fig. 8, time is implicitly
considered as linear. The linear human perception of time is used to guide
the design of the representation of the temporal domain as a horizontal
axis.

F IG . 7 – Diagram representation: a curve represents the structural values.

 F IG . 8 – Visualization process of the diagram representation of Fig. 7.

Fusion at the step “visualization space”. The fusion of the two

visualization processes at the step “visualization space” (Fig. 4-c) implies
that the graphical space of the structural domain depends on the temporal
one. Thus, the characteristics of the graphical space of the temporal
domain and some operators applied on this domain, directly affect the
graphical space of the structural domain, and conversely. We distinguish
two classes of visualization techniques defined by the process of Fig. 4-c
according to whether the graphical space of the structural domain is
merged with the graphical space of the temporal domain, or not.
In the first class, there is one single space representing both domains. If
the temporal domain (respectively the structural domain) is represented in
a three-dimensional space, the structural domain (respectively the
temporal domain) must then be represented in a three-dimensional space.
The spiral representation [2] depicted in Fig. 3-a is an example of
technique that belongs to this group. Fig. 3-a could represent the air
pollution measures of NO (nitric oxide) per month during the period
January 1980 – October 1988. In [25], a spiral representation is used for
visualizing traces of program execution. In both examples, the structural
and temporal domains are visualized into the same graphical space, in this
case, a two-dimensional graphical space. The visualization process of the
representation of Fig. 3-a is presented in Fig. 9: the visualization process
of the structural domain and that of the temporal domain are joined at the
“visualization space” step. In the visualization process of the temporal
domain of Fig. 9 only the cyclic perception of time is considered at the
"point of view on time step. Consequently, time is represented as a spiral.
 F IG . 9 – Visualization process of the spiral representation of Fig. 3-a.

In the second class of techniques, both temporal and structural domains

are represented separately in the same graphical space. For example, our
TimeVis technique of Fig. 10, showing the history of the investments (in
millions of dollars) of a country in agriculture, services and industries,
distinguishes the representation of time from the representation of
structural values. Temporal and structural domains are separately
represented but in the same graphical space. As in the previous example
illustrated in Fig. 3-a, rotating the representation (graphical space) of the
temporal domain (the perspective wall timeline in the example of Fig. 10)
implies a rotation of the representation (graphical space) of the structural
domain.

F IG . 10 – TimeVis technique.

Fusion at the step “point of view on the visualization space”. This

configuration (Fig. 4-d) of the two visualization processes is impossible
because the structural domain is always associated with the temporal one,
which implies that the point of view on the visualization space of the
structural domain is defined only once the point of view on the time space
is defined. Thus, the points of view on the visualization spaces must be
fixed sequentially, considering first the time and then the structural
values. Respecting this ordering constraint we identify two cases depicted
in Fig. 4-e and Fig. 4-f.

F IG . 11 – Superposed Histograms technique [6].

Time and then the step “visualization space of the structural

domain”. Some visualization techniques of temporal data use the
graphical space of the temporal domain as a support for the representation
of the structural domain. The main advantage is the direct association of
structural values with the corresponding temporal values. This direct
association reduces the visual discontinuity between both representations.
As a result, the temporal values and the structural values are visualized in
the same graphical space. First, the characteristics of this graphical space
are all fixed by the temporal domain and second, the structural values are
then visualized within this space. If time is represented in a two-
dimensional space, the structural values will be too. For this reason, the
representation of the structural domain is defined once the visualization
process of time is completely achieved (see Fig. 4-e). It is worth noting
that this configuration is different from that where the fusion proceeds at
the “visualization space” step (see Fig. 4-c). Indeed, based on this
configuration, the visualization process of the temporal domain is fully
independent of that of the structural domain. This latter could be modified
or replaced by another one while keeping the same visualization process
of the temporal domain.

F IG . 12 – Visualization process of the Superposed Histograms technique of

Fig. 11.

The Superposed Histograms Technique (SHT) [6], shown in Fig. 11

illustrates this configuration of the two visualization processes. SHT
visualizes two histories one at a time, allowing the user to switch from
one history to the other. Fig. 11 shows air pollution measures and the user
can switch from NO2 measures to NO ones. In SHT, structural values
have been mapped to 2.5D graphical objects (in the form of cubes). Time
is represented in two dimensions: SHT includes two time-sliders, one
vertical and the other horizontal. Fig. 12 represents the visualization
process (configuration -e- in Fig. 4) of SHT. The visualization process of
the temporal domain is achieved independently from that of the structural
domain, whereas the visualization process of the structural domain
depends on that of the temporal domain. In the visualization process of
the temporal domain, as shown in Fig. 12, two points of view on time are
considered: multi-granular and linear points of view (“point of view on
time” step). The linear point of view on time is used to guide the design
of the perspective wall timeline, whereas the multi-granular point of view
leads us to define two timelines: one, horizontal, to represent hours and
the other, vertical, to represent days. At the “visualization space” step of
the structural domain, a histogram is used to denote the evolution of the
visualized data. The choice of the histogram representation is strongly
influenced by the fact that the data are quantitative. Moreover, histograms
are visualized within a two dimensional graphical space defined by the
visualization process of the temporal domain: this link between the two
visualization processes is represented in Fig. 12 by the arrow from the
“point of view on the time space” to the “structural space” step.
Time and then the step “point of view on the visualization space of
the structural domain”. The process corresponding to this case is shown
in Fig. 4-f. The temporal domain is visualized in a graphical space
independent from the structural domain. On the one hand this distinction
allows an explicit separation between the user tasks referring to each of
the two domains of temporal data, for example by navigating through
time versus navigating along the structural domain. On the other hand,
visual discontinuity may occur, for example making difficult the
identification of a value at a given time.

F IG . 13 – Star Representation technique [8].

 The Star Representation Technique (SRT) [8] of Fig. 13 illustrates the
process of Fig. 4-f.
In the same way as with the Superposed Histograms technique (Fig.
11), SRT supports the visualization of two histories. Fig. 13 shows the
visualization of NO2 concentration measures.

F IG . 14 – Visualization process of the Star Representation technique

of Fig. 13.

Fig. 14 represents its visualization process (configuration -f- in Fig. 4).

The visualization process of the temporal domain of the SRT is achieved
independently from that of the structural domain, and as in SHT the
opposite does not apply. As shown in Fig. 13, SRT relies on two
representations of time used simultaneously (the perspective wall timeline
and the star representation). In the visualization process of the temporal
domain shown in Fig. 14, three points of view on time are considered:
multi-granular, linear and periodic points of view (“point of view on
time” step). The multi-granular point of view leads us to distinguish the
representation of hours from the one of days. The linear point of view on
time is used to visualize days as a perspective wall timeline. The choice
of the star representation (“structural space” step) is strongly influenced
by the fact that we considered the periodic point of view on time: this link
is represented in Fig. 14 by the arrow coming from the “point of view on
time” to the “structural space” step. The star representation could be
visualized within a two-dimensional graphical space as in the example of
Fig. 13 or within a three-dimensional graphical space. In contrast to the
Superposed Histograms technique discussed above (Fig. 11) the structural
space is defined independently from the characteristics of the time space.
As the complete temporal domain visualization process is independent
from the structural domain visualization process in the two configurations
-e- and -f- of Fig. 4, the resulting representation defines a reusable
software technique for interacting with temporal data. For example the
perspective wall timeline software component of the Superposed
Histograms technique (Fig. 11) has been reused for developing the Star
Representation technique (Fig. 13). Moreover we developed a multi-
surface version of the Star Representation Technique by displaying the
perspective wall timeline on a dedicated interaction surface using a PDA
while the star representation is displayed on screen or projected on a wall
[9]. We therefore obtain a generic physical device for interacting with
temporal data.

Process fully separated. Finally the configuration of Fig. 4-g is

impossible as the structural values are always associated with the
temporal values.

5. CONCLUSION AND FUTURE WORK

In this paper, we presented a taxonomy of temporal data visualization
techniques based on the steps of the Chi’s visualization process [4]. Our
taxonomy is useful for systematically exploring the set of possibilities as
well as for thoroughly envisioning visualization techniques of temporal
data. Indeed in [7] we classified thirty-seven visual tools and visualization
techniques of temporal data according to our taxonomy. Most of the
existing techniques fit into the three classes of Fig. 4-a, Fig. 4-b and Fig.
4-c of our taxonomy. It is therefore interesting to further explore and
envision new visualization techniques that would fit in the two other
classes identified in our taxonomy.
Our taxonomy of temporal data visualization techniques is based on
the possible points of contact between two visualization processes, one
dedicated to the temporal domain and one to the structural domain. As
future work, this study could be generalized to any type of data by
considering the visualization of two data spaces or coordinated and
multiple views of the same data space. However, for the case of temporal
data, the two data spaces are linked: indeed the structural values are
stamped by temporal values. For the case of two independent data spaces,
additional configurations should be considered. For instance, the
configuration of Fig. 4-d, which is impossible for temporal data, becomes
possible for two independent data spaces.

6. ACKNOWLEDGMENTS
The work presented in the article is partly funded by the multidisciplinary
French project NAVGRAPHE (as part of a national research program on
large data sets). Many thanks to G. Serghiou for reviewing the paper.

7. REFERENCES
[1] J. Campos, M.-J. Egenhofer and K. Hornsby. Animation model to
Support Exploratory Analysis of Dynamic Environments.
Proceedings of Summer Computer Simulation Conference
(SCSC), pages 761-766, Montreal, Canada, 2003.
[2] J.-V. Carlis and J.-A. Konstan. Interactive Visualisation of Serial
Periodic Data. Proceedings of UIST, pages 29-38, San Francisco,
USA, ACM Press, 1998.
[3] E.-H. Chi. A Taxonomy of Visualization Techniques using the
Data State Reference Model. Proceedings of the IEEE Symposium
on Information Visualization InfoVis, pages 69-76, Salt Lake
City, USA, IEEE Computer Society, 2000.
[4] L. Chittaro and C. Combi. Representation of Temporal Intervals
and Relations: Information Visualization Aspects and their
Evaluation. Proceedings of the 8th International Symposium on
Temporal Representation and Reasoning TIME, pages 13-20,
Cividale del Friuli, Italie, IEEE Computer Society, 2001.
[5] R.-M Church. Evaluation of Quantitative Theories of Timing.
Journal of the experimental analysis of behavior, pages 253-301,
Vol. 71, No 2, Society for the Experimental Analysis of
Behavior, 1999.
[6] C. Daassi, M.-C. Fauvet and L. Nigay. Visual Mining:
Representation Multiplicity of Temporal Data. Proceedings of
DEXA, pages 701-709, Aix-en-Provence, France, LNCS 2453,
Springer, 2002.
[7] C. Daassi. Techniques d’interaction avec un espace de données
temporelles. PhD thesis in Computer Science from Joseph-
Fourier University of Grenoble, France, 2003.
[8] C. Daassi, M.-C. Fauvet and L. Nigay. Visualization process of
Temporal Data. Proceedings of DEXA, pages 914-924, Zaragoza,
Spain, LNCS 3180, Springer, 2004.
[9] C. Daassi, M.-C. Fauvet et L. Nigay. Composant logiciel vs
physique pour interagir avec un espace de valeurs temporelles.
Actes de la conférence francophone sur l'Interaction Homme-
Machine, pages 166-173, Namur, Belgique, ACM Press, 2004.
[10] M. Dumas, C. Daassi, M.-C. Fauvet and L. Nigay. Pointwise
Temporal Object Database Browsing. ECOOP Workshop on
Objects and Databases, pages 170-184, Sophia Antipolis, France,
LNCS 1944, Springer, 2000.
[11] M. Dumas, M.-C. Fauvet, P.-C. Scholl. TEMPOS: A Platform for
Developing Temporal Applications on top of Object DBMS.
Journal on Transactions on Knowledge and Data Engineering,
pages 354-374, Vol. 16, No. 3, IEEE Computer Society, 2004.
[12] S. Havre, B. Hetzler and L. Nowell. ThemeRiver: Visualizing
Theme changes over Time. Proceedings of the IEEE Symposium
on Information Visualization InfoVis, pages 115-123, Salt Lake
City, USA, IEEE Computer Society, 2000.
[13] H. Hochheiser and B. Shneiderman. Using Interactive
Visualizations of WWW Log Data to Characterize Access
Patterns and Inform Site Design. Journal of the American Society
of Information Science and Technology JASIST, pages 331-343,
Vol. 52, No. 4, Wiley, 2001.
[14] J. Ian and A. Wilson. The TimeMap Project: Developing Time-
Based GIS Display for Cultural Data. Journal of GIS in
Archaeology, Vol. 1, ESRI, Redlands, California, 2002.
[15] D.-F. Jerding and J.-T. Stasko. The Information Mural: A
Technique for Displaying and Navigating Large Information
Spaces. Proceedings of the IEEE Symposium on Information
Visualization InfoVis, pages 43-50, Atlanta, USA, IEEE
Computer Society, 1995.
[16] Y. Koike, A. Sugiura and Y. Koseki. TimeSlider: An Interface to
Specify Time Point. Proceedings of UIST, pages 43-44, Alberta,
Canada, ACM Press, 1997.
[17] R. Kosara and S. Miksch. Visualizing Complex Notions of Time.
Proceedings of the Conference on Medical Informatics MedInfo,
pages 211-215, London, UK, IOS Press, 2001.
[18] T. Kramer and J. Jozsa. Visualization and analysis of time-
dependent hydrometric data in windows environment.
Proceedings of the 3rd International Conference on
Hydroinformatics, Copenhagen, Danemarque, A.A. Balkema,
1998.
[19] V. Kumar, R. Furuta and R.-B Allen. Metadata visualization for
digital libraries: interactive timeline editing and review.
Proceedings of the third ACM conference on Digital libraries,
pages 126-133, Pittsburgh, USA, ACM Press, 1998.
[20] A.-M. MacEachren, F.-P Bosco, D. Haug and L.-W. Pickle.
Geographic Visualization: Designing Manipulable Maps for
Exploring Temporally Varying Georeferenced Statistics.
Proceedings of the IEEE Symposium on Information
Visualization InfoVis, pages 87-94, Research Triangle Park, USA,
IEEE Computer Society, 1998.
[21] J.-D Mackinlay, G. Robertson and R. DeLine. Developing
Calendar Visualizers for the Information Visualizer. Proceedings
of UIST, pages 109-119, Marina del Rey, USA, ACM Press,
1994.
[22] W. Müller and H. Schumann. Visualization methods for Time-
Dependent Data - An Overview. Proceedings of the Winter
Simulation Conference WSC, pages 737-745, New Orleans, USA,
ACM Press, 2003.
[23] C. Plaisant, B. Milash, A. Rose, S. Widoff and B. Shneiderman.
LifeLines: Visualizing Personal Histories. Proceedings of
CHI’96, pages 221-227, Vancouver, Canada, ACM Press, 1996.
[24] C.-P. Rainsford and J.-F. Roddick. Visualisation of Temporal
Association Rules. Proceedings of 2nd International Conference
on Intelligent Data Engineering and Automated Learning IDEAL,
pages 91-96, Hong Kong, Chine, LNCS 1983, Springer, 2000.
[25] M. Renieris and S.-P. Reiss. ALMOST: Exploring Program
Traces. Workshop on New Paradigms in Information
Visualization and Manipulation NPIVM, pages 70-77, Kansas
City, USA, ACM Press, 1999.
[26] Y. Shahar and C. Cheng. Intelligent Visualization and
Exploration of Time-Oriented Clinical Data. Topics in Health
Information Management, pages 15-31, Vol. 20, No. 2, Aspen
Publishers,1999.
[27] Y. Shahar and C. Cheng. Model-Based Visualization of Temporal
Abstractions. Computational Intelligence, pages 279-306, Vol.
16, No. 2, Blackwell Publishing, 2000.
[28] B. Shneiderman. The Eyes Have It: A Task by Data Type
Taxonomy for Information Visualizations. Proceedings of the
IEEE Symposium on Visual Languages, pages 336-343, Boulder,
USA, IEEE Computer Society, 1996.
[29] S.-F. Silva and T. Catarci. Visualization of Linear Time-Oriented
Data: a Survey. First International Conference on Web
Information Systems Engineering (WISE), pages 310-319, Hong
Kong, Chine, IEEE Computer Society, 2000.
[30] E.-R Tufte. The Visual Display of Quantitative Information.
Graphics press, 1983.
[31] F. Vernier. La multimodalité en sortie et son application à la
visualisation de grandes quantités d’information. PhD thesis in
Computer Science from Joseph-Fourier University of Grenoble,
France, 2001.
[32] X. Wang, S. Jajodia and V.S. Subrahmanian. Temporal Modules:
an Approach Toward Federated Temporal Databases. Information
Sciences, pages 103-128,Vol. 82, No 1-2, Elsevier Science,1995.