Affiliation Dynamics With An Application To Movie-Actor Biographies

Eurographics/ IEEE-VGTC Symposium on Visualization (2006)
Thomas Ertl, Ken Joy, and Beatriz Santos (Editors)
Affiliation Dynamics with an Application

to Movie-Actor Biographies
Ulrik Brandes, Martin Hoefer † , and Christian Pich
Department of Computer & Information Science, University of Konstanz, Germany
Abstract
We propose a visualization approach for dynamic affiliation networks in which events are characterized by a set
of descriptors. It uses a radial ripple metaphor to display the passing of time and conveys relations among the dif-
ferent constituents through appropriate layout. Our method is particularly suitable when assuming an egocentric
perspective, and we illustrate it on movie-actor biographies.
Categories and Subject Descriptors (according to ACM CCS): information visualization, affiliation networks, time-
dependent visualization
1. Introduction To animate the dynamics of an affiliation network, we use

a concentric ripple metaphor, i.e. events are moving on a ra-
Visual analysis of dynamic data is a challenging task, but of-
dial trajectory. Visual clustering is supported by determining
fers unique possibilities for exploration. With the ever grow-
the angle of a trajectory using barycentric coordinates in a
ing size and complexity of available data, tailored means of
space spanned by the descriptors. In addition, visual clues
visualization for classes of such data are necessary to quickly
about the importance and influence of elements according to
keep track of and identify major changes and patterns in the
statistical or structural indicators are provided. Our visual-
temporal development.
izations thus provide an overview of the dynamics of the col-
We here address the visualization of a particular type of so-
lection of actors, events, and descriptors that constitute the
cial networks, namely affiliation networks [WF94, Ch. 8].
affiliation network. The design is particularly suited for visu-
Affiliation networks are represented as a bipartite graph in
alizing egocentric affiliation networks, i.e. networks whose
which the elements of one vertex set (the actors) are only
scope is defined by those events in which a particular actor
connected to (affiliated with) the elements of the other ver-
(ego) appears.
tex set (the events). It is assumed that the events are time-
stamped and that there is a set of descriptors specifying This article is organized as follows. In Section 2, we pro-
their nature. Examples of such networks are interlocking divide some background and notation for the data model and
rectorates (directors affiliated with company boards distin- its dynamics, and a discussion of related work. The visu-
guished by industry sectors), scientific publications (authors alization design and its layout are described in Section 3.
affiliated with publications characterized by keywords), or An example application of our method to the Internet Movie
terrorist networks (terrorists affiliated with cells character- Data Base is presented in Section 4, and in Section 5 we
ized by activity). We will use a movie database (actors af- conclude with directions for future work.
filiated with movies described by keywords) as our running
example.
2. Affiliation Networks
An affiliation network is represented by a hypergraph N =
† Author acknowledges support by DFG Research Training Group (A, E) consisting of a set A of actors, and a set E of events,
1042 ”Explorative Analysis and Visualization of Large Information where ∅ =6 e ⊆ A for all e ∈ E. If a ∈ e for some a ∈ A and
Spaces e ∈ E, actor a is called affiliated with or incident to event e.

c The Eurographics Association 2006.
U. Brandes, M. Hoefer, and C. Pich / Affiliation Dynamics
We will consider dynamic affiliation networks N = Closest to our approach in terms of visualization charac-
(A, E; desc : E → 2D , T : E → R+ 0 ), in which events are la- teristics is the recent proposal of Appan et al. [AS05] for an
beled using a set D of descriptors and a time-stamp te . Such egocentric layout of instant messenger data. The messages of
a network can, e.g., be represented as a tripartite graph with a single participant are displayed in a circular fashion, where
vertex set V = A ] E ] D as shown in Fig. 1. The following the participant (denoted ego) is placed in the center of the
layout. Nodes representing her communication partners are
A E D located on a circle around it. Egos messages are displayed
e1 te1 as nodes, and are moved from the center outwards to the
a1 d1 corresponding partner. The partner nodes change their size
depending on multitude and content of the communication
with ego. In addition topcial information is displayed. Our
..
. method generalizes this approach in several directions. We
propose a method for the display of general relational data
that is not restricted to chat protocols. Thus, our approach
lacks text-specific elements like the incorporation of topical
ak dm
information. In addition our layout method is more sophis-
el tel
ticated because there is a more complex relational structure
between the two vertex sets in the affiliation network. Events
Figure 1: The tripartite graph with partition sides A (set of
moved in the interior and their relations to the descriptors on
actors), E (set of events) and D (set of descriptors) as well
the circle are of central interest. In the approach of Appan
as time stamps te for every item e ∈ E.
et al., communication partners on the outer circle are most
important, and message nodes are used to simply indicate
notion will be useful to define the similarity vertices, which strength and frequency of communication.
in turn is used to determine their positions. Define the neigh-
borhood Nw of a vertex w ∈ V as the set of vertices incident
to w. Note that, since there are no direct relations between A 3. Radial Layout
and D or within a single set, we have that Na ⊆ E, Nd ⊆ E
and Ne ⊆ A ∪ D for all a ∈ A, e ∈ E and d ∈ D. We use some rather natural intuitions about dynamic affil-
iation networks N = (A, E; desc, T ) to guide our layout de-
sign. We strive to place most recent events prominently in
2.1. Related Work the center of the drawing. Older events are placed more pe-
ripherically, and it should be made clear which events are
Social network visualization has recently been revitalized in
important to attract the center of attention. Older events like
the Nineties [BKR∗ 99,Fre00], and quickly diversifies. Since
descriptors are used to provide a meaningful historical and
affiliation networks, in particular, occur in many disguises,
conceptual context for the new developments. These goals
they are frequently not recognized as such and it is diffi-
are implemented in our layout as follows. At first, fixed po-
cult to give a systematic account of their visualization. A
sitions for the descriptors are chosen on a descriptor circle
straightforward, popular example is a web site called “They
that borders the layout to set up a conceptual context for the
rule” (www.theyrule.net).
events. The time interval under consideration is divided into
In [YFDH01] the authors propose a radial drawing discrete fine-grained ticks, and for each tick a time ripple
method for large networks. The layout is centered around is introduced. Upon introduction, the time ripple is a point
a specified node v, which is placed in the middle of the in the center of the drawing. As time progesses, the ripple
drawing. The graph is then treated as a tree rooted at v. It grows larger in radius, but stays centered in the middle of
is drawn in a radial fashion on concentric circles, through the drawing. An event appearing in the layout remains fixed
which the distance to v becomes apparent. In addition, the to the time ripple corresponding to its time stamp. Hence,
authors describe smooth animation techniques for switch- as time progresses, events appear in the center of the lay-
ing the root node. A radial drawing approach for social out and gradually wander to the outside. The relationships
networks concentrating on structural properties is presented to descriptors and their position on the outer circle then de-
in [BKW03]. In addition there has been a lot of research termine where on the ripple an item is located. Formally, we
on radial space-filling visualization techniques. As an ex- model this using polar coodinates (re , βe ) for event e ∈ E.
ample we mention [SZ00, YMR02], where methods for the The radius re is given by the radius of the time ripple cor-
display of hierarchical information are proposed. A differ- responding to te . The angle βe is determined by the position
ent radial approach for visual correlation is considered in of the descriptors from the neighboorhood Ne of e. In this
[LAMF05]. Visualization of dynamic social networks is dis- way we place most recent events prominently in the center
cussed in [MMB05], and a prototype for visualizing online of the drawing, while older events at the border of the layout
social networks is introduced by [HB05]. provide the historical context.


We apply the HITS algorithm for network analysis to de- Once a tour dσ(1) , . . . , dσ(|D|) , dσ(|D|+1) = dσ(1) with
termine importance and use this value to adjust label sizes |D|
for items and descriptors accordingly. Finally, each actor length |σ| := ∑i=1 δ(dσ(i) , dσ(i+1) ) is found, it is projected
a ∈ A is placed into the context of its neighborhood Na . As to the perimeter of the descriptor circle, or equivalently, to
coordinates we use the average polar coordinates of neigh- the real interval [−π, π), by

boring events.
(
2π
γi−1 + δ dσ(i−1) , dσ(i) |σ| , if 1 < i ≤ |D|,
γi =
−π, otherwise.
3.1. Descriptor ordering
First and foremost the ordering of descriptors on the descrip- 3.2. Radius
tor circle needs to be determined. In order to achieve a mean- The radius function is a fundamental ingredient in the radial
ingful clustering of events, we allocate descriptors using a layout, as it determines the distance of objects to the center.
similarity measure based on the number of common neigh- Furthermore, it is useful for assigning influence values to all
boring events. This is natural, because in turn an event e is events. The events are to pop up in the center at the date of
assigned angle βe according to the neighboring descriptors their time stamp. The movement to the border then decel-
as well. Our descriptor ordering should be designed that later erates with growing age, thus providing recent events with
on events can be placed close to historically and conceptu- more space than older ones. However, events with finite age
ally similar events and descriptors, while dissimilar events never reach the outermost descriptor circle. For these design
are well separated. principles the function
(
In our model, we say that two descriptor nodes d1 and d2 t
if t ≥ 0,
are similar if they have many common neighbors. Therefore, r(t) = t+k
0 otherwise.
we use the distance measure
|Nd1 ∩ Nd2 | is suitable to calculate the radius. The parameter k > 0 deter-
δ(d1 , d2 ) := 1 − 2 mines the time after which an event is in the middle between
|Nd1 | + |Nd2 |
the center and the descriptor circle. More formally, the k de-
to express dissimilarity between two nodes. This index is termines t, for which r(t) = 21 . Fig. 2 shows a plot of the
widely known as the Czekanowski-Dice-Sørensen similar- radius function for some values of k. Given an item e ∈ E,
ity index [CS75] and used for classification and clustering
in fields such as biology or sociology. It is designed such
that 0 ≤ δ(d1 , d2 ) ≤ 1, where δ(d1 , d2 ) = 1 if the two nodes radius function
1
have disjoint neighborhoods and δ(d1 , d2 ) = 0 if the neigh-
borhood is equal. 0.8
For the descriptor circle a cyclic arrangement σ : D → 0.6

radius
{1, . . . , |D|} of all nodes d ∈ D must be found such that the

0.4
sum of distances between consecutive nodes dσ(i) , dσ(i+1) is
minimized. This is equivalent to solving a traveling sales- 0.2 k=1
man problem (TSP) in a complete undirected graph GD = k=2
k=5
(D, D × D), where distances are given by the Czekanowski- 0
0 5 10 15 20 25 30
Dice-Sørensen index. TSP is known to be NP-complete, age of an event
however, since our distance measure δ satisfies the trian-

gle equality (i.e. δ(d1 , d3 ) ≤ δ(d1 , d2 ) + δ(d2 , d3 ) for all
Figure 2: The radius function with k ∈ {1, 2, 5}.
d1 , d2 , d3 ∈ D), our descriptor TSP becomes metric, and
heuristic approximation algorithms for metric TSP can be
used. A simple yet powerful variant is constructing a mini- the radius function determines the ripple with radius r(t −te )
mum spanning tree (MST). A feasible tour is generated by on which e lies at time t. Figure 3 shows ripples for some
walking around the MST, leaving out vertices that have al- multiples of k.
ready been visited. This yields a tour with length at most
twice of the length of the optimal solution. 3.3. Event placement
The best approximation method reaches a factor of The events are located such that the positions are related to
1.5 [Chr79] instead of 2; however, it is slightly more com- the angles of neighboring descriptors. Given a set of k angles
plicated to implement. Both heuristics are known to provide γ1 , . . . , γk , the average angle γ is uniquely determined by
very good near-optimal solutions, so the the hardness of TSP
does not really pose a problem – especially when dealing 1 k 1 k
sin γ = ∑ sin γi , cos γ = ∑ cos γi ,
with instances of moderate size. ρ i=1 ρ i=1

4
r(4k) =
r(∞) = 1
3
5
5
for keyword weighting [BR99] in document indexing. In our
r(3k) = 4 r(5k) =
r(2k) = 2
6
formal framework, the descriptor d ∈ D is weighted with
3
r(k) = 1
2 |D|
wd := log ,
|Nd |
r(0) = 0 giving rarely appearing descriptors more influence than fre-
quent ones that are assumed to be less discriminative.
Using these weights, angles remain fixed once they are
assigned. This may be sufficient in static settings, when our
layout method is employed to display a snapshot of the evo-
lution at a particular point in time. In the dynamic context,
Figure 3: Time ripple with various radii.
however, we want descriptors characterizing older events to
be less influential than descriptors adjacent to recent events.
To this end, we introduce the notion of temporal influence of
where an event e at time t dependent on the age of a event. As the
v
u !2 !2 radius is a convenient indicator of age, and the radius func-
u k k tion r has the desirable property to map to [0, 1], we express
ρ=t ∑ sin γi + ∑ cos γi . the influence of event e as 1 − r(t − te ). The dynamic weight
i=1 i=1
of a descriptor d then becomes
It can be obtained by computing
 ! ∑ (1 − r(t − te ))
 1 k k
wd (t) := log
e∈E
.
arccos ρ ∑ cos γi , if ∑ sin γi ≥ 0,

(1 − r(t − te ))


i=1 i=1
∑
θ= ! e∈Nd
 1 k
− arccos ρ ∑ cos γi , otherwise.



i=1
3.4. Importance
See Figure 4 for an example of an average angle. A straight-
When visualizing large data sets there is typically the need
forward way of placing an event is to use the average angle
of providing more important or interesting items with more
of all its descriptors. However, we incorporate information
space than others. Therefore, it is necessary to be able to
about time and distinctness into the contribution of descrip-
express importance for events and descriptors, and to cre-
tor angles. The average angle is biased with weights wd for
ate a meaningful space assignment by assigning font sizes.
each descriptor d.
A widely-known approach is mutual reinforcement for im-
π portance computation. A node in a graph has a high impor-
2
γ1 tance score if its neighbors are considered important. We use
γ2
this approach for the bipartite graph of events and descrip-
γ
tors. Events are considered important if they are associated
with many important descriptors, and in turn, descriptors are
π given a high score if the events they describe are important.
−π 0
Let x ∈ [0, 1]E and y ∈ [0, 1]D denote importance vec-
tors. The importance xe for each item e ∈ E and yd for
γ3 descriptor d ∈ D can be found by iteratively computing
xe ← ∑d∈Ne yd for each e ∈ E and yd ← ∑e∈Nd xe for each
d ∈ D and and normalizing x and y to unit length by di-
− π2 1/2
viding each entry by the Euclidean length ∑e∈E xe2 or
Figure 4: The average angle γ of γ1 , γ2 , γ3 . 1/2
∑d∈D y2d , respectively. This is done iteratively until the
maximum change of importance for any node in the itera-
To improve cluttering properties of the layout it is de- tion falls below a given threshold ε > 0. It is the essence
sirable to give discriminative descriptors more weight than of the well-known HITS algorithm [Kle99] for calculating
others. This moves an event e into the direction of descrip- hubs and authorities. In our layout these importance values
tors of Ne that are less frequently present in other neighbor- are used to scale the font sizes of events and descriptors. In
hoods, thus giving e a more distinct direction. This idea has order to adjust the space occupied by the label, we use the
been formulated in the area of information retrieval using square root of the values multiplied with a constant to derive
the notion of inverse document frequency. It was introduced the exact font sizes.

3.5. Actor placement whether it is possible to divide works and collaborators into
meaningful groups. Hence, we use an ego-centric approach
Actors are to be placed into the layout in a structured man-
focusing on the biography of a single actor (denoted as ego).
ner. It is important to support the visual presentation of clus-
tering properties and temporal development of the affiliation For an ego we obtained all movies he has participated
network. Hence, an actor should always be located close to in with co-starring actors, and extracted keywords of the
the important adjacent events that are present in the layout at movies in the IMDB. A tripartite actor-movie-keyword
the current point in time. We thus locate an actor a in a po- graph was set up, where movies are events, and keywords
lar barycenter of the events in an adjusted neighborhood as are descriptors. The time stamp for a movie is the year, in
follows. Let Na (t) be the present neighborhood at time t, i.e. which it appeared in the cinemas. For this graph we use the
the set of all adjacent events e with time stamp te ≥ t. At time steps outlined above to create a keyword graph, from which
t actor a then gets located at polar coordinates (ra (t), αa (t)), we derive positions of keywords on the descriptor circle us-
where the radius is derived by the average age of the present ing the MST heuristic for TSP.
neighborhood: In the layout ego is located in the center and releases
movies over time. Thus, in the course of the animation, the
!
1
ra (t) = r ∑ t − te .
|Na (t)| e∈N
names of all the ego’s movies pop up in the center and mono-
a(t) tonically wander to the outside of a circular area, while their
For the angle αa a weighted average angle is used (see label size depends on their importance. Using weighted av-
above). At time t we determine the weighted average angle erage angles, movies eventually cluster in similarity groups
of events in the present neighborhood Na (t). The weights are and show a topical and temporal clustering of an actor’s bi-
given by the normalized authority values associated with the ography. Actors are placed with weighted polar average co-
events in Na (t). ordinates in order to locate them close to the movies they
appeared in. This forms a visual clustering of movies and
actors in the conceptual space generated by keywords, and
3.6. Animation provides an overview of the development of an actors per-
sonal biography.
In general we cannot assume a uniform distribution of time
stamps. In addition, there are important applications, in Figures 5 and 6 on the next pages show our layout for ac-
which time stamps are rather coarse-grained (e.g. publica- tor John Travolta in 1999. The graph consists of 5044 nodes
tion year). Our layout, however, should embed the course of – 3435 actors, 78 movies and 1531 keywords. For the lay-
events continuously. Hence, we need methods to distribute out we used k = 10 in the radius computation. Keywords
positioning changes onto the set of fine grained ticks in or- d used only by one movie (with |Nd | = 1) were omitted
der to provide a smooth and stable animation. because they are irrelevant for identifying topical connec-
tions of movies. Furthermore, there were simply too many
For presentation, we will assume that time stamps are
keywords to include them into a readable, moderately-sized
given as years. We consider the positions of an event e in two
overview. After this adjustment 251 keywords remained in
consecutive years t and t + 1 as key frames. For the remain-
the graph. For the same reasons we only display actors that
ing ticks we interpolate between the two positions. Thus, at
have appeared in at least three movies up to the current time
time t + τ (0 ≤ τ < 1) the polar coordinates for an event e are
t, i.e. for which |Na (t)| ≥ 3.

r(t + τ), τ · βe (t + 1) + (1 − τ) · βe (t). The IMDB contains entries for award shows and TV ap-
This results in a polar interpolation of the key frame po- pearances that are identified as movies. Our method can pro-
sitions. The event is moved circularly to its new position, vide a visual filter for these items clustering them on a hor-
which is natural in the given context, whereas a direct izontal line in the right part of Fig. 5. In addition, observe
straight line through the inner of the circle would not comply that the earlier dancing and music movies cluster in the up-
with the overall goals of the layout. For actors the animation per right region. Later movies are located in the lower part
works accordingly. of the image. Fig. 6 provides a closeup view on the music
cluster, where correspondence of keywords, movies and ac-
tors can be verified. Furthermore importance values seem
4. Application: Actor Biographies to roughly correspond to commonly perceived influence of
In this section we apply our method to a data set com- Travoltas movies.
ing from the internet movie data base (IMDB, imdb.com). Fig. 7 shows screenshots of an example animation de-
When exploring such huge datasets there is often interest in picting the actor biography of Arnold Schwarzenegger. The
focusing on the curriculum vitae of one particular individ- graph consists of 4991 nodes – 3650 actors, 74 movies and
ual. Typically, one wants to know which movies have been 1267 keywords. Again, we removed keyword nodes of de-
most important for this actor. Likewise, the similarity of two gree one leaving 173 keywords in the graph. The ”Termina-
given movies might be of interest, clustering properties, or tor” blockbusters and related items cluster in the lower part

of the images. In addition, they are also assigned the largest References
font sizes, which correctly mirrors their perceived societal [AS05] A PPAN P., S UNDARAM H.: Egocentric Analysis
influence. Award shows appear on a horizontal line in the and visualization of instant messaging activity. Tech. Rep.
right half of the drawing, together with actors that co-appear AME-TR-2005-19, Arizona State University, 2005.
with Arnold Schwarzenegger only in award shows and in no
other movies. [BDJ99] B ERRY M., D RMAC Z., J ESSUP E.: Matrices,
vector spaces and information retrieval. SIAM Review 40
Due to the increasing amount of events our layout is likely (1999), 335–361.
to experience visual cluttering in the border region. This is
supported by the decreasing distance between time ripples. [BKR∗ 99] B RANDES U., K ENIS P., R AAB J., S CHNEI -
DER V., WAGNER D.: Explorations into the visualization
Here one can use dynamic importance values that decrease if
the movies have passed a certain age. This results in smaller of policy networks. Journal of Theoretical Politics 11, 1
font sizes when the label reaches the border region. Another (1999), 75–106.
alternative would be to restrict the layout to the display of [BKW03] B RANDES U., K ENIS P., WAGNER D.: Com-
events from a certain time interval (e.g. for each time t to the municating centrality in policy network drawings. IEEE
interval [t − 10,t]). Transactions on Visualization and Computer Graphics 9,
2 (2003), 241–253.
5. Conclusion [BR99] BAEZA -YATES R., R IBEIRO -N ETO B.: Modern
Information Retrieval. ACM Press, 1999.
We have presented a novel layout method to display tempo-
ral development of tripartite relations derived from affiliation [Chr79] C HRISTOFIDES N.: Worst-case analysis of a new
networks. A classic application of the method lies in the area heuristic for the traveling salesman problem. Tech. Rep.
of information retrieval and uses the vector-space model for 388, Grad. School of Industrial Admin., CMU, 1979.
document analysis [BDJ99]. In a collection of documents [CS75] C LIFFORD H., S TEPHENSON W.: Introduction
(events) each document contains a number of relevant terms into Numerical Classification. Academic Press, 1975.
(descriptors). It is natural to assume that documents have
[Fre00] F REEMAN L.: Visualizing social networks. Jour-
authors or publishers (actors), and that for each document
nal of Social Structure 1, 1 (2000).
temporal information is available about the time of writing.
Using our visualization it is possible to observe the devel- [HB05] H EER J., B OYD D.: Vizster: visualizing online
opment of the collection in terms of topical structure or fre- social networks. In Proc. IEEE Information Visualization
quency of document creation. Furthermore, publishers and (2005), pp. 32–39.
authors can be located in the topical environment of their [Kle99] K LEINBERG J.: Authoritative sources in a hyper-
documents. As a special case recent studies have focused linked environment. Journal of the ACM 46, 5 (1999),
on the analysis of bibliographic networks and co-authorship. 604–632.
Here our method can be used to visualize the existence and [LAMF05] L IVNAT Y., AGUTTER J., M OON S.,
structure of research communities. In addition, it is possi- F ORESTI S.: Visual correlation for situational awareness.
ble to pick journals or authors as sources and keywords or In Proc. IEEE Information Visualization (2005), p. 13.
journals as descriptors to get a topical or a publisher-related
picture of the work under consideration. [MMB05] M OODY J., M C FARLAND D., B ENDER -
DE M OLL S.: Visualizing network dynamics. American
There are a number of directions, in which our method Journal of Sociology 110, 4 (2005), 1206–1241.
can be extended. An interesting adjustment is to derive a dy-
[SZ00] S TASKO J., Z HANG E.: Focus+context display
namic importance measure for the events. The definition of
and navigation techniques for enhancing radial, space-
present neighborhood can be extended to the whole network,
filling hierarchy visualization. In Proc. IEEE Information
i.e. we define the current network at time t as the tripartite
Visualization (2000), pp. 57–65.
subgraph induced by event nodes e with time stamp te ≥ t.
This network can then be used with the HITS algorithm cal- [WF94] WASSERMAN S., FAUST K.: Social Network
culating importance values for the present time t. Analysis: Methods and Applications. Cambridge Univer-
sity Press, 1994.
Another adjustment would be to include interactive ele-
ments like mouse-over effects. For instance, a mouse-over [YFDH01] Y EE K.-P., F ISHER D., D HAMIJA R.,
effect for an actor, event or keyword could be used to identify H EARST M.: Animated exploration of dynamic graphs
and highlight the neighborhood of this item. Also, methods with radial layout. In Proc. IEEE Information Visualiza-
to move the layout freely in time or to interactively zoom can tion (2001), pp. 43–50.
be included. Finally, it is possible to extend our visualiza- [YMR02] YANG J., M.WARD , RUNDENSTEINER E.: In-
tion into three dimensions. This, however, poses some new terring: An interactive tool for visually navigating and ma-
challenges when generating a good keyword distribution, as nipulating hierarchical structures. In Proc. IEEE Informa-
there is no easy reduction to TSP anymore. tion Visualization (2002), pp. 77–84.

onship
Father Son Relati
end
f Fri
th O
Relationship
t er
Dea
bus
Sexma,
ge
et
n Scene
taellodra
dy, Intern
Mother Son
Words, Defecatio
atherrria
k
Love
y
Child Custo
Prison
Family
Bloht c
Pregnanc
ari, M
el
Childbirth
Christmas
Adultery
r
Park
Docto
Taxicab
Sequ
Play On
ne
r
Kids And
, Siste
molic
Airpla
ck
Baby,
co
Wife
Fig
ndtra
Bro
M and,
ath
Dis
PCre
rk
Husb
w y
Yo
Ne Cit
Sou
nd
ba
0s
lyn rk
Hus
197
e
l
Fir
ok Yo
ita
r
Ba
Of
Narration
Broew
ss
nia
sp
Lo
ip
N
ylva
sh
ns
Ho
Pen
do nd
m
r
Dance
t
on ie
en
CFr
e
e c y
an
us cid
cu
es m
Propaganda
Ro
alo Ac
,R or
ve
rn
op
Je ar ay ntest
Go
oft
Ro
C bw Co
y
dit
Su nce
y
Nu
r End ship rve

illa
nce
ac
ns
pir
Co
Daee
T
n
rtdIneThteer Relation Su s
50
e 19
rad
Pa
nce
as
Tex
u
M is ugh Da ticle
Twther Da Base
dO
n Ar
Fa Courtroom
n
uatio
, Grad
ertible
Town
Small
Conv
h
Beac ol
Scho
Cult
High
Book
Science,
Diner
In Film
Illness
Terminal
TV Show
Tragedy
FBI
Wheelchair
Based On True Story

TV Special
Awards Show
Celebrity
Live Performance
Award, Oscar
Cigarette Smoking
Insurance
Scam
Black Co Hollywood
medy
Los A
Basketb ngeles
Witnes all, Wig Califo
rnia
s
Under
Split age
Scree Drinki
n ng
P
Titro
lefaSnity
R pok
Prod
ucer
Film
Gaacia en B
Mak
Nig yS l S
yC
htcr lur
ing Film lu
Ne In Film Obs ha racte
oN Film
cene
Fing
er
lub r
ma
Actr kin
g
Ges
oir ess
Behind
The
Scenes
ture
Sm
8 T okin
g ra Go
ld
FTaoplessck Ex
Sp
y
To
plo ll ile
tS
ce
Kn
din F igh
t
ne
SShhot g ro
Sh m
Co nife
VRuap
oInt Th ip
K
ca In
He
lge
Fa
Ine Ch
in T
ar
ily
, In
e, hig
ity
igh
Miam
Ind
sa
nity
Tehst
Sh h
i Fl
Kidn
appi
or
ng
ida
Eth las
ep
C
t
ot
e
C
hil
Alco me er, Di
en
nic s D
In
Doepfath
Le
St
dM
Rac
ho stic vorc
RDeru
Po ll Bat
Th
de
l
Slu iffe
Base
ism
Sw
g
urd
e
Pois
Wed
nt
lice
S
r
Sh
Fun
Ba
er
Cem
ba
Vio e
Expl
vegns de, Sho
itch
h
College
F
Drug ge
on
Ho
c
Helico
oding
Professor
ding
Violenc
le
etery
ilm
o
Heavy Rain
k
Based On
Urination Scene
Flashback Sequence
Gun
Cellular Phone
sta
nce
o
ren
bla
Car
t tTIn
Enforc
eral
ge t In Th
ionship
ce
o
ement
s
DT
Nove
eh
ae
pter
th For
e
Hea e
ehe
Blo d,
ad
od
SplM
Father Son Relat
artia
l Art
s
d
ien
f Fr
Figure 5: A snspshot of John Travolta’s movie biography layout in 1999. Dancing and music movies are concentrated in the
th O
Relationship
ster
upper right region, while his later movies tend downwards.

ea
e x a, D
kbu
age
aelloSdram
rne t
e
cation Scen
Mother Son
stody, Inte
Words, Defe
Love
ncy
Prison
ly
ri
Bloht c
m c, M
d Fami
el
Child Cu
t
Pregna
Childbirth
Christmas
Mroatherr
Adultery
r
i
Docto
r
Kids Anab
Sequ
Play On
Taxicne
Sister
Par
a
Airpla
ack
Baby,
PCraetholi
Wife,
Sou co
Fig
and,
ndtr
Dis
rk
Husb
B
ew ity
Yo
nd
nN C
a
s
sb
kly ork
1970
e
l
Hu
Fir
ita
oo Y
Ba
Of
Narration
Br ew
ss
ia
p
Lo
p
n
N
lva
hi
os
sy
nn
m s
Pe
do nd
H
cer
Dan
t
on e
en
CFri
e ce y
an
us id
cu
lo cc
es m
da
Propagan
Ro
or
,R rn
a A
e
p ov
Je ar ay ntest
fto o
G
Ro
C ubw
y
dit o
Nu
S eC
ncn
y
ac
e pir
Daee
nc ns
Co
illa
Su
rve
e
T 1 95
0s
d ra
Pa
nce
s
xa
Te
Da Art
icle
T On
ed
Bas
Courtroom
ation
radu
le, G
vertib
Con
Beach ol
Scho
High
Diner
Figure 6: A closeup view of John Travolta’s dancing movie region, as indicated by some characteristic keywords. The font TV Show
In Film
size illustrates that and Saturday Night Fever is associated with more important keywords than its sequel, Staying Alive. Due to Wheelchair
co-appearance in those movies, Olivia Newton-John is placed to the same region, as well.
TV Special
Awards Show
Live Performance
Award, Oscar
Hollywood

Los A
Bask ngeles
etball,
W ig Califo
Under
age
rnia
Drinki
ng
P
Titrlofanit
R e Spoyk
Pr
Gaacia
Nigy Slurl Slur en By C
Fil
Fil
N Ob
s
ht cen
ha clu ract
Film
mak
ing
eF
ing
er b Ge
er
tre
Ac
stu
re
ss
Behind
The
Scenes
Sm
8 T okin
Go
To rack ld
Ex
p le
ss
Fa Sp
y
To
p ll lo
ile
tS
ce
Kn
F din ig
ht
n e
SShhot g ro
Sh
Co nife
VRa
oInt Th ip
ul pe
K
c ai In
ga
Fa
Ine Ch
m
ne
ily
r
, In
ity
Mia
sa
Ind
,S
nit
Test
mi
y
Kid
Flo
na
pp
h
in

the lower region of the image “Terminator” movies are clustered.
Figure 7: Actor biography of Arnold Schwarzenegger. The snapshots are taken in intervals of five years, starting in 1982. In
2007 2002
ity
ity
Nud e
Nud e
enc
enc
Android
Android
llig
llig
Post
Post
onFemale
onFemale
Travel
Travel
ry,
Stati
Stati
em
em
ntary,
Tea
Tea
, Gasial Inte
, Gasial Inte
Tim
Tim
arityific
arityific
r,AT
ce
r,AT
ence menta
, SW
, SW
ip
ip
quenial Comme
e
e
dity, Art
dityle, Art
g, Soc
sene,sVulg
r Wa
sene,sVulg
r Wa
nsh
nsh
Sequ ial Com
Stor
or
or
e Ey
e Ey
ll
ll
tio
tio
Nu
Th
Th
lasGu
lasGu
ck Se
ent Stor
el Favorite
el Favorite
leapse
leapse
g Ma
g Ma
ela
ela
d In
d In
Samplin
ship
ship
le Nu le
leent
back g, Soc
el To Cult
el To Cult
ngne
rR
rR
chi
chine
oppin
oppin
shba
te
te
Making Of
Behind The Scenes
Making Of
Behind The Scenes
The Terminator
Special Effects
The Terminator
Special Effects
Cyborg
Cyborg
Film Making
Film Making
Fatale
Fatale
Interview
il
Interview
lation
lation
Remake
Remake
Neo Noir
Neo Noir
Nuccaly
Nuccaly
Apo
Apo
Motorcyc
Motorcyc
, Sh
, Sh
Sequ
Sequ
Ma
Ma
Dep
Departm
Sequ
Sequ
Suartm
Sung
Ma
Ma
gh
gh
Voi
Voice
Flace
Flash Samplin
Femme
Femme
Sex
Sex
e
e
n Re
n Re
Hardcore
Hardcore
ileP, eril
au
au
Psychotronic
Psychotronic
un ay Mirr
un ay Mirr
down
down
Pool
Pool
In
In
Reveng
Reveng
Tw
Tw
ma
ma
Loveo W
Loveo W
Swimming
Swimming
ow
ow
G
G
r D Stabbe
r D Stabbe
mob
mob
Aviation
Aviation
Hu
Hu
rism
Sh
Sh
told
told
Children
Children
Terro
Terro
the
Teen
the ile, er
Teen
Chi
Chirism
Au
Au
Arm d
Arm d
Journalism
Journalism
e g lou
e g lou
rance
rance
Fa
Fa
Hollywood
Hollywood
Appea
Appea
The LeC
Journalist
The LeC
Journalist
n n
Cameo
Cameo
io io
Historical
Historical
st st
In Thom
In Thom
Pun
Pun
ss ss
Family
Family
And
And
Kids
Kids
ott In ro
ott In ro
pa pa
Boy
Boy
MCrane
MCrane
ocau ocau
Girl,
Girl,
SShhoush
SShhoush
Hol Hol
Toy
Com
Toy
ar Com
s,
ar
s,
Fire
Fire
es es
Political
Political
l
l
Anima
Anima
ucle ucle
Election Election
Party
dn Party
dn
Democratic Democratic
eN eN
Drugs Kin Drugs Kin
Homelessness Homelessness
Party Party
Fac
Convention
Political
Republican
Food
kse Fac
Convention
Political
Republican
Food
ks e
Protest Protest
The uca The uca

rica rica
Ame Ame
Latin Latin
edy edy
t In t In
Monkey Monkey
Com k, Com k,
Shoc Shoc
dy
TCr h dy
TCr h
tric Box tric Box
Sho un Sho un
Elec Elec
Bud Bud
st one st one
r ns r, r ns r,
Che Chef
eph Che Chef
eph
The Tel gG Drain The Tel gG Drain
t In t In
ale
povatoent ale
povatoent
Sho Sho
urode urode
tlin shipStorm tlin shipStorm
n r eE n r eE
s Msi we W ,
m s Msi we W ,
m
M asoolo
toutroaretal lace
GSkaull htlationgeles M asoolo
toutroaretal lace
GSkaull htlationgeles
Exp le M P
ecTh
id uct Sh
Flam
u rsign ReLos An
se So ,
Exp le M P
ecTh
id uct Sh
Flam
u rsign ReLos An
se So ,
NLiqu d
rnia
herrFigure
LMaotthe ce rnia
NLiqu d
rnia
herrFigure
LMaotthe nce rnia
lifo Fa lifo Fa
PCaro Ca
Pro
ty
Par
C len alifo Par
ty
le C alifo
t
ure
Fut
Vio chbla
eles de Fut
ure
t
Vio chbla
eles de
iden
Acc Swit Ang Acc
iden Swit Ang
Car Car
LoersSon Relationship LoersSon Relationship
Fath Fath
Fiction
Non
Fiction
Non
e e
War PollericRobot War PollericRobot
Car Chase Kil cider Killed, SelfCarSacrifice Car Chase Kil cider Killed, SelfCarSacrifice
rming Robot Sui On
Office ing Back rming Robot Sui On
Office ing Back
Transfo Police LandThe Transfo Police LandThe
rian Body
Shot In rian Body
Shot In
Veterina Broken Leg Body Veterina Broken Leg Body
California
San Francisco
ing
Explod California
San Francisco
ing
Explod
To You
Happy Birthday
Exp loding Car

Helicopter To You
Happy Birthday
Exp loding Car

Helicopter
Bear Bear
Shot To Death Shot To Death
Blood Splatter Blood Splatter
Awards Show Awards Show
Shot In The Forehead Shot In The Forehead
TV Special TV Special
Urination Scene Urination Scene
Undercover
Person On Fire
Shot In The Head Undercover
Person On Fire
Shot In The Head
Premarital Sex Premarital Sex
Throat Slitting Throat Slitting
Humiliation Humiliation
Camel, Wizard Camel, Wizard
Based On Based On
Pulp Magazine Pulp Magazine
Stabbed Stabbed
In The Chest In The Chest
Decap Decap
Villainess Villainess
Fight itation Fight itation
Obscene Obscene
Finger Finger
Gesture Gesture
Moral Cult Fa Moral Cult Fa
Amb Amb
Indepe Indepe
Redem iguity, Resp vorite Redem iguity, Resp vorite
ndent ndent
Film Film
Ma n ect, Hono Ma n ect, Hono
r r
rtiaptio rtiaptio
Based Based
Battle, Battle,
On Comic On Comic
l Ar l Ar
Book Book
Sac
Mask Mask
rific
ts Sacrific ts
e, Wa e, Wa
rrior rrior
Barbarian Barbarian
Sword Sword
Sword Fight Sword Fight
Woman Woman
Book
On
Based
And Book
On
Based
And
Sorcery Sorcery
, Tra , Tra
Actre Actre
gedy
Weightlifting
gedy
Weightlifting
ss ss
Go Go
Building
Body
re Building
Body
re
Poetic Poetic
Justic Justic
e e
Based Based
On On
Novel Im Novel Im
Cg Sev
ere
p ale Cg Sev
ere
pa
d Arm d Arm
Ki Ki
Friei Friei lem
Co dn Co dn
me en
M ndship Tw nverappi nt M ndship Tw nverappi t
ist tible ng ist tible ng
Tra In Tra In
Con in Con in
Im X spir Im X spir
Ho Ra acy Ho Ra acy
Th Th
BBlolourd BBlolourd
Wa sta ye Wa sta ye
pe pe
Taxi shi ge d Taxi shi ge d
rs eE rs eE
cab ngt Sk cab ngt Sk
on ele on ele
ocker ocker
Nec nd Nec nd
Sur D.C Sur D.C
at at
Hit ton Hit ton
on d b
k
on d b
rea . rea k .
l In l In
ion ion
Re Cr Re Cr
Ne
Bre Ne
Bre
w ve otc ak w ve otc ak
us us
Yo rse h Yo rse h
Exp
Exp
rk
Fo in rk
Fo in
lodi
lodi
ota g ota g
ng
ng
Be
Be
ge ge
tra
tra
W
W
Hou
Hou
ter ter
Sce Sce
Fat
Fat
se,
se,
ya
ya
l
l
Child
Child
Space
Space
ne ne
Son her
Son her
Des
Des
ea
Psych
, Zoo
ea
Psych
, Zoo
ert
ert
Mars,
Mars,
ologic
ologic
Fa
Fa
Eag
al
Eag
al
False
False
le
le
po
Holog
po
Holog
Dram
Dram
Memo
Memo
a
Yel
a
Yel
Car
Car
ry,
ram
ry,
ram
Nudity
Nudity
ll Fr
ll Fr
n
n
ling
ling
Rebel
Rebel
Inventor
Inventor
lion
lion
om
om
Actor
Actor
, Uns
, Uns
Mansion
Mansion
Playing
ubt
Playing
ubt
He
He
Cyberp
Cyberp
unk
unk
itled
itled
Multiple
Multiple
Fu
Fu
Roles
Roles
De
De
ight
ight
For
For
Racism
Racism
Tit
Tit
eign
eign
ath
ath
Lan
Lan
turis
turis
Brea
Brea
Sword
Sword
Actor
Actor
Director
Director
gua
gua
sts
sts
tic
tic
le Sp
le Sp
T800
T800
ge
ge
Millennium
Millennium
Back
Back
From
From
T10
T10
Christmas
Christmas
The
The
oken
oken
Wea
Wea
00
00
Dead
Dead
Superhero
Superhero
Crew
Crew
Cast
Cast
By
By
Slavery
Slavery
Predator
Predator
Co
Co
Monster
Monster
Alien
Alien
pons
Psych
pons
Psych
Dream
Dream
Robot
Teacher, Villain
Robot
Teacher, Villain
Ch
Ch
Hero
Hero
Moura
Moura
Trap, Hunter, Jungle

opath
opath
arac
arac
Of Mas
Of Mas
Warrior
Warrior
ter
ter
rphge
rphge
ing
ing
s Des
s Des
truction
truction
1997 1992
ity
ity
roid
Nud e
Nud e
enc
enc
Android
t And
llig
llig
Post
Pos
onFemale
onFemale
Travel
Travel
Stati
Stati
em
em
ntary,
ntary,
Tea
Tea
, Gasial Inte
, Gasial Inte
Tim
Tim
arityific
arityific
r,AT
ce
r,AT
ial
ence me
, SW
, SW
ip
ip
quen Comme
e
e
dity, Art
dityle, Art
g, Soc
sene,sVulg
r Wa
sene,sVulg
r Wa
nsh
nsh
Sequ ial Com
Stor
or
e Ey
e Ey
ll
ll
irror
Nu
Th
Th
lasGu
lasGu
ck Se
ent Stor
el Favorite
el Favorite
latio
latio
leapse
leapse
M
d In
d In
Samplin
ship
ship
le Nu le
leent
back g, Soc
Re
Re
el To Cult
el To Cult
ngne
chi
chine
op
op
shba
ter
ter
Making Of
Behind The Scenes
Making Of
Behind The Scenes
The Terminator
Special Effects
The Terminator
Special Effects
Cyborg
Cyborg
Film Making
Film Making
Fatale
Fatale
Interview
Interview
lation
lation
Remake
Remake
Way
Neo Noir
Neo Noir
Nuccaly
Nuccaly
Apo
Apo
Motorcyc
Motorcyc
, Sh
, Sh
Sequ
Sequ
Ma
Ma
Dep
Departm
Sequ
Sequ
Suartm
Sung
Ma
Ma
gh
gh
Voi
Voice
, il
, il
Flace
Flash Samplin
Femme
Femme
n
Sex
Sex
e
e
n Re
n Re
Hardcore
Hardcore
au
au
Psychotronic
Psychotronic
un ay Mirr
down
down
Pool
Pool
Reveng
Reveng
Tw
Tw
ma
ma
Loveo W
Lo o
Swimming
Swimming
ow
ow
d
d
G
Guve
r D Stabbe
r D Stabbe
Aviation
Aviation
Hu
Hu
rism
Sh
Sh
told
told
Children
Children
Terro
Terro
rm
rm
the
Teen
the
Teen
Chi
Chirism
Au
Au
A
A
Journalism
Journalism
e g lou
e g lou
rance
rance
Fa mobInilePer ping Ma
Fa mobInilePer ping Ma
Hollywood
Hollywood
Appea
Appea
The LeC
Journalist
The LeC
Journalist
n n
Cameo
Cameo
io io
Historical
Historical
st st
In Thom
In Thom
Pun
Pun
ss ss
Family
Family
And
And
Kids
Kids
ott In ro
ott In ro
pa pa
Boy
Boy
MCrane
MCrane
ocau ocau
Girl,
Girl,
SShhoush
SShhoush
Hol Hol
Toy
Com
Toy
ar Com
s,
ar
s,
Fire
Fire
es es
Political
Political
l
l
Anima
Anima
ucle ucle
Election Election
Party
dn Party
dn
eN eN
Drugs Kin Drugs Kin
Party Party
Fac
Convention
Political
Republican
Food
kse Fac
Convention
Political
Republican
Food
ks e
Protest Protest
The uca The uca

rica rica
Ame Ame
Latin Latin
edy edy
t In t In
Monkey Monkey
Com k, Com k,
Shoc Shoc
dy
TCr h dy
TCr h
tric Box tric Box
Sho un Sho un
Elec Elec
Bud Bud
st one st one
r ns r, r ns r,
Che Chef
eph Che Chef
eph
t In t In
ale
povatoent ale
povatoent
Sho Sho
urode urode
n r eE n r eE
s Msi we W ,
m s Msi we W ,
m
M asoolo
toutroaretal lace
toutroaretal lace
Exp le M P
ecTh
id uct Sh
Flam
u rsign ReLos An
se So ,
Exp le M P
ecTh
id uct Sh
Flam
u rsign ReLos An
se So ,
NLiqu d
rnia
herrFigure
LMaotthe ce rnia
NLiqu d herrFigure
LMaotthe nce rnia
lifo Fa lifo Fa
PCaro Prornia Ca
ty
Par
C len alifo ty
Par
le C alifo
t
ure
Fut
Vio chbla
eles de Fut
ure
t
Vio chbla
eles de
iden
Acc Swit Ang Acc
iden Swit Ang
Car Car
Fath Fath
Fiction
Non
Fiction
Non
e e
rming Robot Sui On
Office ing Back
rian Body
Shot In rian Body
Shot In
California
San Francisco
ing
Explod California
San Francisco
ing
Explod
To You
Happy Birthday
Exp loding Car

Helicopter To You
Happy Birthday
Exp loding Car

Helicopter
Bear Bear
Undercover
Person On Fire
Person On Fire
Shot In The Head
Based On Based On
Stabbed Stabbed
Decap Decap
Obscene Obscene
Finger Finger
Gesture Gesture
Amb Amb
Redem iguity, Resp
Indepe
vorite Redem iguity, Resp
Indepe
vor ite
ndent ndent
Film Film
r r
rtiaptio rtiaptio
Based Based
Battle, Battle,
On Comic On Comic
l Ar l Ar
Book Book
Sac
Mask Mask
rific
ts Sacrific ts
Barbarian Barbarian
Sword Sword
Book
On
Based
And Book
On
Based
And
Sorcery Sorcery
, Tra , Tra
ged ged
Actre e, Warrior Woman Actre e, Warrior Woman
y
Weightlifting
y
Weightlifting
ss ss
Go Go
Building
Body
re Building
Body
re
Poetic Poetic
Justic Justic
e e
Based Based
On On
Novel Im Novel Im
Cg Sev
ere
p ale Cg Sev
ere
pa
d Arm d Arm
Ki Ki
Friei Friei lem
Co dn nd Co dn
me sh en
M ndship Tw nverappi nt M ip Tw nverappi t
Tra In Tra In
Con in Con in
Im X spir Im X spir
Ho Ra acy Ho Ra acy
Th Th
BBlolourd BBlolourd
Wa sta ye Wa sta ye
pe pe
rs eE rs eE
on cab ngt Sk cab ngt Sk
on ele on ele
ocker
Nec nd Nec nd
Sur D.C Sur D.C
at at
Hit ton Hit ton
ocdkebr
k
on d b
rea . rea k .
l In l In
ion ion
Re Cr Re Cr
Ne
Bre Ne
Bre
us us
Yo rse h Yo rse h
Exp
Exp
rk
Fo in rk
Fo in
lodi
lodi
ota g ota g
ng
ng
Be
Be
ge ge
tra
tra
W
W
Hou
Hou
ter ter
Sce Sce
Fat
Fat
se,
se,
ya
ya
l
l
Child
Child
Space
Space
ne ne
Son her
Son her
Des
Des
ea
Psych
, Zoo
ea
Psych
, Zoo
ert
ert
Mars,
Mars,
ologic
ologic
Fa
Fa
Eag
al
Eag
al
False
False
le
le
po
Holog
po
Holog
Dram
Dram
Memo
Memo
a
Yel
a
Yel
Car
Car
ry,
ram
ry,
ram
Nudity
Nudity
ll Fr
ll Fr
n
n
ling
ling
Rebel
Rebel
Inventor
Inventor
lion
lion
om
om
Actor
Actor
, Uns
, Uns
Mansion
Mansion
Playing
ubt
Playing
ubt
He
He
Cyberp
Cyberp
unk
unk
itled
itled
Multiple
Multiple
Fu
Fu
Roles
Roles
De
De
ight
ight
For
For
Racism
Racism
Tit
Tit
eign
eign
ath
ath
Lan
Lan
turis
turis
Brea
Brea
Sword
Sword
Actor
Actor
Director
Director
gua
gua
sts
sts
tic
tic
le Sp
le Sp
T800
T800
ge
ge
Millennium
Millennium
Back
Back
From
From
T10
T10
Christmas
Christmas
The
The
oken
oken
Wea
Wea
00
00
Dead
Dead
Superhero
Superhero
Crew
Crew
Cast
Cast
By
By
Slavery
Slavery
Predator
Predator
Co
Co
Monster
Monster
Alien
Alien
pons
Psych
pons
Psych
Dream
Dream
Robot
Teacher, Villain
Robot
Teacher, Villain
Ch
Ch
Hero
Hero
Moura
Moura

opath
opath
arac
arac
Of Mas
Of Mas
Warrior
Warrior
ter
ter
rphge
rphge
ing
ing
s Des
s Des
truction
truction
1987 1982
ity
ity
roid
roid
Nud e
Nud e
enc
enc
t And
t And
llig
llig
Pos
onFemale
onFemale
Travel
Travel
ry, Pos
Stati
Stati
em
em
nta
ntary,
Tea
Tea
, Gasial Inte
, Gasial Inte
Tim
Tim
arityific
arityific
r,AT
ce
r,AT
ial
ence me
, SW
, SW
ip
ip
quen Comme
e
e
dity, Art
dityle, Art
g, Soc
sene,sVulg
r Wa
sene,sVulg
r Wa
nsh
nsh
Sequ ial Com
Stor
or
e Ey
e Ey
ll
ll
tio
tio
irror
Nu
Th
Th
lasGu
lasGu
ck Se
ent Stor
el Favorite
el Favorite
leapse
leapse
g Ma
g Ma
M
ela
ela
d In
d In
Samplin
ship
ship
le Nu le
leent
back g, Soc
el To Cult
el To Cult
ngne
rR
rR
chi
chine
oppin
oppin
shba
te
te
Making Of
Behind The Scenes
Making Of
Behind The Scenes
The Terminator
Special Effects
The Terminator
Special Effects
Cyborg
Cyborg
Film Making
Film Making
Fatale
Fatale
Interview
il
Interview
lation
lation
Remake
Remake
Way
Neo Noir
Neo Noir
Nuccaly
Nuccaly
Apo
Apo
Motorcyc
Motorcyc
, Sh
, Sh
Sequ
Sequ
Ma
Ma
Dep
Departm
Sequ
Sequ
Suartm
Sung
Ma
Ma
gh
gh
Voi
Voice
Flace
Flash Samplin
Femme
Femme
n
Sex
Sex
e
e
n Re
n Re
Hardcore
Hardcore
ile, eril
au
au
Psychotronic
Psychotronic
un ay Mirr
down
down
Pool
Pool
In
Reveng
Reveng
Tw
Tw
ma
ma
ob P
Loveo W
Lo o
Swimming
Swimming
ow
ow
G
Guve
r D Stabbe
r D Stabbe
mob
ld
Aviation
Aviation
Hu
Hu
rism
Sh
Sh
tom In
told
Children
Children
Terro
Terro
the
Teen
the ile, er
Teen
Chi
Chirism
Au
Au
Arm d
Arm d
Journalism
Journalism
e g lou
e g lou
rance
rance
Fa
Fa
Hollywood
Hollywood
Appea
Appea
The LeC
Journalist
The LeC
Journalist
n n
Cameo
Cameo
io io
Historical
Historical
st st
In Thom
In Thom
Pun
Pun
ss ss
Family
Family
And
And
Kids
Kids
ott In ro
ott In ro
pa pa
Boy
Boy
MCrane
MCrane
ocau ocau
Girl,
Girl,
SShhoush
SShhoush
Hol Hol
Toy
Com
Toy
ar Com
s,
ar
s,
Fire
Fire
es es
Political
Political
l
l
Anima
Anima
ucle ucle
Election Election
Party
dn Party
dn
eN eN
Drugs Kin Drugs Kin
Party Party
Fac
Convention
Political
Republican
Food
cks
e Fac
Convention
Political
Republican
Food
cks
e
Protest Protest
The ua The
rica ru a rica
Ame Ame
Latin Latin
edy edy
t In t In
Monkey Monkey
Com k, Com k,
Shoc Shoc
dy
TCr h dy
TCh
tric Box tric Box
Sho un Sho un
Elec Elec
Bud Bud
st one st one
r ns r, r ns r,
Che Chef
eph Che Chef
eph
t In t In
ale
povatoent ale
povatoent
Sho Sho
urode urode
n r eE n r eE
s Msi we W ,
m s Msi we W ,
m
M asoolo
toutroaretal lace
toutroaretal lace
Exp le M P
ecTh
id uct Sh
Flam
u rsign ReLos An
se So ,
Exp le M P
ecTh
id uct Sh
Flam
u rsign ReLos An
se So ,
NLiqu d
rnia
herrFigure
LMaotthe ce rnia
NLiqu d herrFigure
LMaotthe nce rnia
lifo Fa lifo Fa
PCaro Prornia Ca
ty
Par
C len alifo ty
Par
le C alifo
t
ure
Fut
Vio chbla
eles de Fut
ure
t
Vio chbla
eles de
iden
Acc Swit Ang Acc
iden Swit Ang
Car Car
Fath Fath
Fiction
Non
Fiction
Non
e e
rming Robot Sui On
Office ing Back
rian Body
Shot In rian Body
Shot In
California
San Francisco
ing
Explod California
San Francisco
ing
Explod
To You
Happy Birthday
Exp loding Car

Helicopter To You
Happy Birthday
Exp loding Car

Helicopter
Bear Bear
Undercover
Person On Fire
Person On Fire
Shot In The Head
Based On Based On
Stabbed Stabbed
Decap Decap
Obscene Obscene
Finger Finger
Gesture Gesture
Amb Amb
Redem iguity, Resp
Indepe
vorite Redem iguity, Resp
Indepe
vor ite
ndent ndent
Film Film
r r
rtiaptio rtiaptio
Based Based
Battle, Battle,
On Comic On Comic
l Ar l Ar
Book Book
Sac
Mask Mask
rific
ts Sacrific ts
e, Wa e, Wa
rrior rrior
Barbarian Barbarian
Sword Sword
Woman Woman
Book
On
Based
And Book
On
Based
And
Sorcery Sorcery
, Tra , Tra
Actre Actre
gedy
Weightlifting
gedy
Weightlifting
ss ss
Go Go
Building
Body
re Building
Body
re
Poetic Poetic
Justic Justic
e e
Based Based
On On
Novel Im Novel Im
Cg Sev
ere
p ale Cg Sev
ere
pa
d Arm d Arm
Ki Ki
Friei Friei lem
Co dn nd Co dn
me sh en
M ndship Tw nverappi nt M ip Tw nverappi t
Tra In Tra In
Con in Con in
Im X spir Im X spir
Ho Ra acy Ho Ra acy
Th Th
BBlolourd BBlolourd
Wa sta ye Wa sta ye
pe pe
rs eE rs eE
on cab ngt Sk cab ngt Sk
on ele on ele
ocker
Nec nd Nec nd
Sur D.C Sur D.C
at at
Hit ton Hit ton
ocdkebr
k
on d b
rea . rea k .
l In l In
ion ion
Re Cr Re Cr
Ne
Bre Ne
Bre
us us
Yo rse h Yo rse h
Exp
Exp
rk
Fo in rk
Fo in
lodi
lodi
ota g ota g
ng
ng
Be
Be
ge ge
tra
tra
W
W
Hou
Hou
ter ter
Sce Sce
Fat
Fat
se,
se,
ya
ya
l
l
Child
Child
Space
Space
ne ne
Son her
Son her
Des
Des
ea
Psych
, Zoo
ea
Psych
, Zoo
ert
ert
Mars,
Mars,
ologic
ologic
Fa
Fa
Eag
al
Eag
al
False
False
le
le
po
Holog
po
Holog
Dram
Dram
Memo
Memo
a
Yel
a
Yel
Car
Car
ry,
ram
ry,
ram
Nudity
Nudity
n
n
ling
ling
Rebel
Rebel
Inventor
Inventor
lion
lion
om
om
Actor
Actor
, Uns
, Uns
Mansion
Mansion
Playing
Playing
He
He
Cyberp
Cyberp
unk
unk
ll Fr ubtitled
ll Fr ubtitled
Multiple
Multiple
Fu
Fu
Roles
Roles
De
De
ight
ight
For
For
Racism
Racism
Tit
Tit
eign
eign
ath
ath
Lan
Lan
turis
turis
Brea
Brea
Sword
Sword
Actor
Actor
Director
Director
gua
gua
sts
sts
tic
tic
le Sp
le Sp
T800
T800
ge
ge
Millennium
Millennium
Back
Back
From
From
T10
T10
Christmas
Christmas
The
The
oken
oken
Wea
Wea
00
00
Dead
Dead
Superhero
Superhero
Crew
Crew
Cast
Cast
By
By
Slavery
Slavery
Predator
Predator
Co
Co
Monster
Monster
Alien
Alien
pons
Psych
pons
Psych
Dream
Dream
Robot
Teacher, Villain
Robot
Teacher, Villain
Ch
Ch
Hero
Hero
Of
Of
Moura
Moura

opath
opath
arac
arac
Mas
Mas
Warrior
Warrior
ter
ter
rphge
rphge
ing
ing
s Des
s Des
truction
truction

Affiliation Dynamics With An Application To Movie-Actor Biographies

Cargado por

Información del documento

Descripción original:

Título original

Derechos de autor

Formatos disponibles

Compartir este documento

Compartir o incrustar documentos

Opciones para compartir

¿Le pareció útil este documento?

¿Este contenido es inapropiado?

Copyright:

Formatos disponibles

Affiliation Dynamics With An Application To Movie-Actor Biographies

Cargado por

Copyright:

Formatos disponibles

Eurographics/ IEEE-VGTC Symposium on Visualization (2006)

Thomas Ertl, Ken Joy, and Beatriz Santos (Editors)

Affiliation Dynamics with an Application

Ulrik Brandes, Martin Hoefer † , and Christian Pich

Department of Computer & Information Science, University of Konstanz, Germany

1. Introduction To animate the dynamics of an affiliation network, we use

For the descriptor circle a cyclic arrangement σ : D → 0.6

{1, . . . , |D|} of all nodes d ∈ D must be found such that the

however, since our distance measure δ satisfies the trian-

r End ship rve

Based On True Story

upper right region, while his later movies tend downwards.

The uca The uca

Exp loding Car

Exp loding Car

Trap, Hunter, Jungle

The uca The uca

Exp loding Car

Exp loding Car

Trap, Hunter, Jungle

Exp loding Car

Exp loding Car

Trap, Hunter, Jungle

También podría gustarte