Information visualization and text mining: application to a corpus on posthumanism

The world we live in generates so much data that the very
structure of our societies has been transformed as a result.
The way we produce, manage and treat information is changing;
and the way we present this information and make it available
raises numerous questions. Is it possible to give information
an intuitive form, one that will enable understanding and
memory? Can we give data a form better suited to human
cognition? How can information display enhance the utility,
interest and necessary features of data, while aiding memories
to take hold of it? This article will explore the relevance
of extracting and visualizing data within a corpus of text
documents revolving around the theme of posthumanism.
When a large quantity of data must be represented, visualizing
the information implies a process of synthesis, which can be
assisted by techniques of automated text document analysis.
Analysis of the text documents is followed by a process of
visualization and abstraction, which allows a global visual
metaphoric representation of information. In this article, we
will discuss techniques of data mining and the process involved
in creating a prototype information-cartography software; and
suggest how information-cartography allows a more intuitive
exploration of the main themes of a textual corpus and
contributes to information visualization enhancement.
Introduction
Human civilization is now confronted with the problem of
information overload. How can we absorb an ever-increasing
quantity of information? How can we both comprehend it and
fi lter it according to our cognitive tools (which have adapted
to understand general forms and structures but struggle with
sequential and cumulative data)? How can we mine the riches
that are buried in information? How can we extract the useful,
necessary, and essential pieces of information from huge
collections of data?
Most importantly, how can we guarantee the literal survival of
memory? For why should we remember when an uncountable
number of machines remember for us? And how can we
remember when machines won’t allow us to forget? Human
memory has endured through time precisely because it is
unaware of the greater part of the signals the brain receives
(our senses gather over 11 million pieces of information per
second, of which the brain is only aware of a maximum of 40
(Philipps, 2006, p.32)). Memory exists because it can forget.
Is an electronic archive a memory? We must rethink the way
that we produce and handle information. Can we give it a more
intuitive form that would better lend itself to retention and
understanding? Can we better adapt it to human cognition?
How can we extrapolate narratives from databases? How
can we insert our memories into this machine memory? The
answer is simple: by visualizing it.
For the past few years, the technical aspect of information
representation and visualization has been the subject of active
research which is gaining more and more attention (Card et
al., 1999; Chen, 1999; Don et al., 2007; Geroimenko and Chen,
2005; Perer and Shneiderman, 2006; Spence, 2007). Despite
the richness of the work that has been done, there is still a
glaring lack of projects related to textual analyses, specifi cally of
literary or theoretical texts, which have successfully integrated
advances in the fi eld of information visualization.
Objectives
Our project is part of this visual analytical effort. How can
we visually represent posthumanism? How can we produce an
image of its questions and challenges? How can we transform
the enormous quantity of associated information the concept
carries into something intuitive? We believe that the creation
of a thematic map is part of the solution. Why a map? Because
the information-driving suffocation we experience is also
our inability to create cognitive maps (as Fredric Jameson
suggested). And so we challenged ourselves to create a map
of posthumanism, one that could be read and understood
intuitively.
To do this, we have chosen to use Google Earth (GE) as
the basic interface for the project. GE’s interface is naturally
intuitive. It corresponds to our collective imagination and it also
allows users to choose the density of information they desire.
GE allows the user to “dive” into the planet’s geographical,
political, and social layers, or to stay on higher, more general
levels. GE “tells” the story of our way of seeing the world. GE
shows us the world the way science describes it and political
maps draw it.
This leads us to confront three primary challenges: the fi rst
was to fi nd an area to represent posthumanism. The second
was to give this area a visual form. The third was to integrate,
in an intuitive way, the signifi cant number of data and texts on
the subject. Methodology
In order to successfully create this thematic map, we fi rst
compiled a signifi cant number of texts about posthumanism.
The methodology used to treat the corpus was inspired by
previous work in the fi eld of text mining (Forest, 2006; Ibekwe-
Sanjuan, 2007; Weiss et al., 2005). The goal of the analysis is to
facilitate the extraction and organization of thematic groups.
Non-supervised clustering technique is used in order to
produce a synthetic view of the thematic area of the subject
being treated. The data is processed in four main steps:
1. Segmentation of the documents, lexical extraction and fi ltering
2. Text transformation using vector space model
3. Segment classifi cation
4. Information visualization
Having established and sorted a sizeable informationpopulation,
a new continent was needed. The continental
outline of Antarctica was adopted. Antarctica was chosen
because its contours are relatively unknown and generally
unrecognizable; it tends to be thought of more as a white sliver
at the bottom of the world map than a real place. Politically
neutral Antarctica, whose shape is curiously similar to that of
the human brain, is a large area surrounded by oceans. These
qualities made it symbolically ideal for utilization in a map of
posthumanism, a New World of the 21st century.
Antarctica also allowed us to avoid information overload typical
of a populated area: it carries minimal associative memories
and historic bias; few people have lived there. To differentiate
our new continent, the contour of Antarctica was moved into
the mid Pacifi c. The continent was then reskinned visually with
terrain suggestive of neurology, cybernetics, and symbiosis. The
end result was a new land mass free of memories ready for an
abstract population.
Visualizing the results
After identifying the thematic structure of the corpus, themes
are converted into regions. Themes are assigned portions of
the continent proportional to their occurrence. Inside each
major region are 2 added levels of analysis: sub-regions and subsub-
regions, each represented by several keywords. Keywords
correspond to clusters discovered during the automated text
analysis. So it is possible to burrow physically downward into
the subject with greater and greater accuracy; or to rest at an
attitude above the subject and consider the overall ‘terrain’.
Each area is associated with a category of documents. From
each area it is possible to consult the documents associated
with each region.
Figure 1. A general level of the visualized thematic map
The system offers users a certain number of areas, based on
the algorithms used to process the data. These areas represent
themes. Clicking on the brain icons allows the user to read an
excerpt from one of the texts that is part of the cluster.
Figure2. A specifi c level of the thematic map
When the user zooms in on a region, the application shows
the next level in the hierarchy of data visualization. Within one
theme (as shown below) several sub-themes appear (in red).
A greater number of excerpts is available for the same area. Figure 3. Looking at a document from the thematic map
The icons indicating the availability of texts may be clicked on
at any time, allowing the user to read the excerpt in a small
pop-up window, which includes a link to the whole article. This
pop-up window can serve to show pertinent images or other
hyperlinks.
Figure 4. 3-dimensional rendering of the thematic map
At any time, the user can rotate the point of view or change its
vertical orientation. This puts the camera closer to the ground
and allows the user to see a three dimensional landscape.
Conclusion
We see the visualization of data, textual or otherwise, as part
of a fundamental challenge: how to transform information into
knowledge and understanding. It is apparent that the signifi cant
amount of data produced by research in both science and the
humanities is often much too great for any one individual.
This overload of information sometimes leads to social
disinvestment as the data eventually cancel each other out.
We think that giving these data an intuitive form will make
their meaning more understandable and provide for their
penetration into the collective consciousness. Posthumanism
seems particularly well adapted to pioneer this system because
it questions the very defi nition of what it is to be human.
References
Alpaydin, E. (2004). Introduction to machine learning. Cambridge
(Mass.): MIT Press.
Aubert, N. (2004). “Que sommes-nous devenus?” Sciences
humaines, dossier l’individu hypermoderne, no 154, novembre
2004, pp. 36-41.
Card, S. K., Mackinlay, J. and Shneiderman, B. (1999). Readings
in information visualization: using vision to think. Morgan
Kaufmann.
Chen, C. (1999). Information visualisation and virtual
environments. New York: Springer-Verlag.
Clement, T., Auvil, L., Plaisant, C., Pape, G. et Goren, V. (2007).
“Something that is interesting is interesting them: using text
mining and visualizations to aid interpreting repetition in
Gertrude Steins The Making of Americans.” Digital Humanities
2007. University of Illinois, Urbana-Champaign, June 2007.
Don, A., Zheleva, E., Gregory, M., Tarkan, S., Auvil, L., Clement,
T., Shneiderman, B. and Plaisant, C. (2007). Discovering
interesting usage patterns in text collections: integrating text
mining with visualization. Rapport technique HCIL-2007-08,
Human-Computer Interaction Lab, University of Maryland.
Garreau, J. (2005). Radical Evolution. New York: Doubleday.
Geroimenko, V. and Chen, C. (2005). Visualizing the semantic
web: XML-based Internet and information visualization. New
York: Springer-Verlag.
Forest, D. (2006). Application de techniques de forage de
textes de nature prédictive et exploratoire à des fi ns de gestion
et d’analyse thématique de documents textuels non structurés.
Thèse de doctorat, Montréal, Université du Québec à
Montréal.
Frenay, R. (2006). Pulse, The coming of age of systems, machines
inspired by living things. New York: Farrar, Strauss and Giroux.
Harraway, D. (1991). Simians, cyborgs, and women. The
reinvention of nature. New York: Routledge.
Ibekwe-SanHuan, F. (2007). Fouille de textes. Méthodes, outils et
applications. Paris: Hermès.
Jameson, F. (1991). Postmodernism, or, the cultural logic of late
capitalism. Verso. Joy, B. (2000). “Why the future doesn’t need us.” Wired. No
8.04, April 2000.
Kelly, K. (1994). Out of control. The rise of neo-biological
civilization. Readings (Mass.): Addison-Wesley Publishing
Company.
Kurzweil, R. (2005). The singularity is near: when humans
transcend biology. Viking.
Levy, S. (1992). Artifi cial life. New York: Vintage Books, Random
House.
Manning, C. D. and H. Schütze. (1999). Foundations of statistical
natural language processing. Cambridge (Mass.): MIT Press.
Perer, A. and Shneiderman, B. (2006). “Balancing systematic
and fl exible exploration of social networks.” IEEE Transactions
on Visualization and Computer Graphics. Vol. 12, no 5, pp. 693-
700.
Philips, H. (2006). “Everyday fairytales.” New Scientist, 7-13
October 2006.
Plaisant, C., Rose, J., Auvil, L., Yu, B., Kirschenbaum, M., Nell
Smith, M., Clement, T. and Lord, G. (2006). “Exploring erotics
in Emily Dickinson’s correspondence with text mining and
visual interfaces.” Joint conference on digital libraries 2006.
Chapel Hill, NC, June 2006.
Ruecker, S. 2006. “Experimental interfaces involving visual
grouping during browsing.” Partnership: the Canadian journal
of library and information practice and research. Vol. 1, no 1, pp.
1-14.
Salton, G. (1989). Automatic Text Processing. Reading (Mass.):
Addison-Wesley.
Spence, R. (2007). Information visualization: design for interaction.
Prentice Hall.
Tattersall, I. (2006). “How we came to be human.” Scientifi c
American, special edition: becoming human. Vol 16, no 2, pp.
66-73.
Unsworth, J. (2005). New methods for humanities research.
The 2005 Lyman award lecture, National Humanities Center,
11 November 2005.
Unsworth, J. (2004). “Forms of attention: digital humanities
beyond representation.” The face of text: computer-assisted text
analysis in the humanities. The third conference of the Canadian
symposium on text analysis (CaSTA). McMaster University, 19-
21 November 2004.
Weiss, S. M., Indurkhya, N., Zhang, T. and Damereau, F. J.
(2005). Text mining. Predictive methods for analyzing unstructured
information. Berlin; New York: Springer-Verlag.