EFFECTIVE ORGANIZATION AND VISUALIZATION OF WEB SEARCH
RESULTS
Nicolas Bonnel
IRISA
Rennes, France
nicolas.bonnel@irisa.fr
Vincent Lemaire and Alexandre Cotarmanac’h
France Telecom, Division R&D
Lannion and Issy-Les-Moulineaux, France
vincent.lemaire@francetelecom.com
alexandre.cotarmanach@francetelecom.com
Annie Morin
IRISA
Rennes, France
amorin@irisa.fr
ABSTRACT
While searching the web, the user is often confronted by a
great number of results, generally displayed in a list which
is sorted according to the relevance of the results. Facing
the limits of this approach, we propose to explore new or-
ganizations and presentations of search results, as well as
new types of interactions with the results to make their ex-
ploration more intuitive and efficient. The main topic of
this paper is the processing of the results coming from an
information retrieval system. Although the relevance de-
pends on the results quality, the effectiveness of the results
processing represents an alternative way to improve the rel-
evance for the user. Given the current expectations, this
processing is composed by an organization step and a vi-
sualization step. Then the proposed approach organizes the
results according to their meaning using a Kohonen Self-
Organizing Map (SOM), and visualizes them in a 3D scene
to increase the representation space. The 3D metaphor pro-
posed here is a city.
KEY WORDS
Search Results Visualization, 3D Metaphors, Human-
Computer Interfaces, Web Mining, Evaluation.
1 Introduction
Searching the web is one of the most frequent tasks, but
often one of the most frustrating too. Search engines which
are a way to represent the web to the users, are mainly
used for web searches. However they are as easy to use as
their results are difficult to interpret, which shows the web
search contradiction. Indeed it becomes more and more
difficult to extract the relevant information for a given
search since available data on the World Wide Web is con-
stantly increasing. The search engines return a number of
results so great as it is necessary to search for new methods
to process these results. These methods must be more
adapted thanks to: a more relevant result organization, a
richer visualization interface and an intuitive navigation in
the result space.
This paper deals with the processing of query results.
This processing, still neglected in some information re-
trieval systems, is becoming more and more important and
essential. It can be considered as a solution for enriching
the results. It is in fact complementary to the search
process and is also a way to increase the result “relevance”
for the user. If the result quality remains a major concern,
the quality of the result restitution (organization and
visualization) must be taken into account too. Without an
effective organization of the results, the user has to process
manually the huge amount of results or refine the query in
order to limit the results. This last solution can be com-
pared to use a search engine for searching into the results!
So these alternative solutions require efforts from the users.
Facing the increase of query results, it seems natural
to want to organize and visualize them in an effective
and adapted way. That explains the goal of the presented
approach, which is to offer the user a search interface
enabling him to quickly find the relevant information. The
two main points to reach this goal are a good document
organization and an effective visualization. Concerning
these two aspects, our directions are a clustering method
(the self-organizing maps) and a 3D visualization. The
choice of a 3D visualization enables to exploit cognitive
metaphors (and more specially the spatial metaphors) in a
more intuitive way. 3D offers new interaction possibilities
too. So it enables us to bring a new point of view to the
result visualization. However, considerable new prob-
lems appear, such as the navigation in such an environment.
This paper deals with the unsupervised organization
of documents and the graphical representation of the re-
sults. A previous work [1] already deals with the same
processing of web search results (i.e. a SOM-based orga-
nization of the results and a 3D city-based visualization),
but it is mainly intended to give a description of our proto-
type. In this paper we focus more precisely on the SOM-
based organization and the 3D visualization based on the
city metaphor, as two independent tasks and without taking
the prototype implementation into account. And an evalua-
tion of this 3D metaphor is also proposed. Considering the
above introduction, this paper is structured as follows. The
next section proposes a short overview of the main issues
in search results processing. Then Section 3 explains the
self-organization method and Section 4 is devoted to the
visualization. The last section allows us to conclude and
209516-054

gives an outlook on future work.
2 Two Issues in Search Results Processing
The search results processing has two main issues: results
clustering and results visualization. For the first one, the
goal is to find an effective method which allows to group
similar results together and to organize the various clusters.
The second one is to find an effective visualization of the
organized results. These two points are briefly discussed in
the following subsections.
2.1 Search Results Visualization
Many works have been done on search results visualiza-
tion in the last few years (some examples can be found
in [2] [3]). The goal of this subsection is not to present
an exhaustive overview of the various approaches in this
field, but only to give some information to locate our
visualization approach (proposed in Section 4) in the
literature and to explain our choices.
In this paper, one particularity is that the visualiza-
tion is done on organized (or clustered) search results.
Therefore the considered methods are only those which
show the content-based links between the documents. An-
other constraint on the visualization is that the document
projection shows the semantic proximity. The taxonomy
of search results visualization systems proposed in [1]
enables to show where the methods of visualization which
address this problem are located in the literature. As this
kind of visualization groups similar documents together,
it gives the user information on the next document to select.
Visualization of inter-document similarities requires
at least two dimensions to be efficient. So the techniques
used in this approach have already given up the linear
display of ordered lists. Indeed the two main techniques
are graphs and maps. The meta search engine KARTOO1
proposes a cartography of search results but a drawback
is the lack of an overview of these results. Another
well-known example is the WEBSOM2 project [4]. The
map approach can take advantage of the cognitive aspect.
That is why geographic metaphors are often used such
as in MAP.NET3 (developed by ANTARCTICA) or in [5].
However with the increase of the results and the links
complexity, graphs and maps become more and more
unreadable.
So one idea is to exploit 3D visualizations to increase
the available space to represent information. The added di-
mension allows the display of complex graphs in a more
1http://www.kartoo.com
2http://websom.hut.fi/websom
3http://maps.map.net
readable way. This third dimension can also be used for re-
placing maps by 3D worlds: landscapes [6] or cities [7] [8].
Other works are the AVE method [9] and its PERISCOPE
system which are the closest works to those ones presented
in this paper. We have the use of mixed interfaces (3D
scene and 2D interface) in common, or the use of many vi-
sualization metaphors which answer different goals. How-
ever the approach proposed in this paper takes the prob-
lem of data organization in a “semantic” point of view into
account. Indeed it is not sufficient in the context of web
search to only order the pages according to some low-level
descriptors. However the dimension increase makes navi-
gation essential and especially more complex. We are fac-
ing another problem which is not obvious to solve.
2.2 Search Results Clustering
There are many works in the area of text classification
which is not discussed in this paper. Indeed our goal
is more to find a good projection of documents and/or
clusters than the classification itself.
Clustering can be used to improve retrieval results,
which has been investigated in many previous works
[10] [11]. The two main possibilities are static cluster-
ing (pre-clustering on the entire corpus) and on-the-fly
clustering which can be considered as a post-retrieval
document browsing technique (e.g. the clustering engine
VIVI´SIMO4). In this paper, we are only interested in
the second solution. Moreover the context of this paper
imposes to have an unsupervised method for organizing
the documents. Among techniques which address this
problem, one of them is particularly interesting: the
Kohonen self-organizing maps [12]. Indeed this method
enables to cluster and to project documents onto an output
space (generally a 2D space). In other words, it is a
clustering method which organizes documents (or word
vectors) on a map with predefined size, which guarantees
a good use of space during the visualization.
Moreover the obtained organization has a neighbor-
hood concept. Indeed two neighboring documents on the
map have similar word vectors. Privileged application ar-
eas of the SOM are visualization and cluster analysis [13].
With the SOM, it is also possible to have hierarchy levels
or a map with dynamic size [14]. These two points can be
interesting in our context but are not exploited at this time.
Self-organizing maps have already been used for textual
data clustering such as in the WEBSOM project [4] or
in [15].
However the organization of web search results im-
plies a particular SOM application whose adaptation is de-
scribed in subsection 3.1. For example, web search results
are special textual documents due to their various size, con-
4http://www.vivisimo.com
210