A Brief Survey of Text Mining

Andreas Hotho
KDE Group
University of Kassel
[email protected]
Andreas Nürnberger
Information Retrieval Group
School of Computer Science
Otto-von-Guericke-University Magdeburg
[email protected]
Gerhard Paaß
Fraunhofer AiS
Knowledge Discovery Group
Sankt Augustin
[email protected]
May 13, 2005

Abstract
The enormous amount of information stored in unstructured texts cannot sim-
ply be used for further processing by computers, which typically handle text as
simple sequences of character strings. Therefore, specific (pre-)processing meth-
ods and algorithms are required in order to extract useful patterns. Text mining
refers generally to the process of extracting interesting information and knowledge
from unstructured text. In this article, we discuss text mining as a young and in-
terdisciplinary field in the intersection of the related areas information retrieval,
machine learning, statistics, computational linguistics and especially data mining.
We describe the main analysis tasks preprocessing, classification, clustering, in-
formation extraction and visualization. In addition, we briefly discuss a number of
successful applications of text mining.

1 Introduction
As computer networks become the backbones of science and economy enormous quan-
tities of machine readable documents become available. There are estimates that 85%
of business information lives in the form of text [TMS05]. Unfortunately, the usual
logic-based programming paradigm has great difficulties in capturing the fuzzy and

1
often ambiguous relations in text documents. Text mining aims at disclosing the con-
cealed information by means of methods which on the one hand are able to cope with
the large number of words and structures in natural language and on the other hand
allow to handle vagueness, uncertainty and fuzziness.
In this paper we describe text mining as a truly interdisciplinary method drawing
on information retrieval, machine learning, statistics, computational linguistics and es-
pecially data mining. We first give a short sketch of these methods and then define
text mining in relation to them. Later sections survey state of the art approaches for
the main analysis tasks preprocessing, classification, clustering, information extraction
and visualization. The last section exemplifies text mining in the context of a number
of successful applications.

1.1 Knowledge Discovery

In literature we can find different definitions of the terms knowledge discovery or
knowledge discovery in databases (KDD) and data mining. In order to distinguish
data mining from KDD we define KDD according to Fayyad as follows [FPSS96]:

”Knowledge Discovery in Databases (KDD) is the non-trivial process of

identifying valid, novel, potentially useful, and ultimately understandable
patterns in data”

The analysis of data in KDD aims at finding hidden patterns and connections in
these data. By data we understand a quantity of facts, which can be, for instance, data in
a database, but also data in a simple text file. Characteristics that can be used to measure
the quality of the patterns found in the data are the comprehensibility for humans,
validity in the context of given statistic measures, novelty and usefulness. Furthermore,
different methods are able to discover not only new patterns but to produce at the same
time generalized models which represent the found connections. In this context, the
expression “potentially useful” means that the samples to be found for an application
generate a benefit for the user. Thus the definition couples knowledge discovery with a
specific application.
Knowledge discovery in databases is a process that is defined by several processing
steps that have to be applied to a data set of interest in order to extract useful patterns.
These steps have to be performed iteratively and several steps usually require interac-
tive feedback from a user. As defined by the CRoss Industry Standard Process for Data
Mining (Crisp DM1 ) model [cri99] the main steps are: (1) business understanding2 , (2)
data understanding, (3) data preparation, (4) modelling, (5) evaluation, (6) deployment
(cf. fig. 13 ). Besides the initial problem of analyzing and understanding the overall
task (first two steps) one of the most time consuming steps is data preparation. This
is especially of interest for text mining which needs special preprocessing methods to
1 http://www.crisp-dm.org/
2 Business understanding could be defined as understanding the problem we need to solve. In the context

of text mining, for example, that we are looking for groups of similar documents in a given document
collection.
3 figure is taken from http://www.crisp-dm.org/Process/index.htm

2
 Figure 1: Phases of Crisp DM

convert textual data into a format which is suitable for data mining algorithms. The ap-
plication of data mining algorithms in the modelling step, the evaluation of the obtained
model and the deployment of the application (if necessary) are closing the process cy-
cle. Here the modelling step is of main interest as text mining frequently requires the
development of new or the adaptation of existing algorithms.

1.2 Data Mining, Machine Learning and Statistical Learning

Research in the area of data mining and knowledge discovery is still in a state of great
flux. One indicator for this is the sometimes confusing use of terms. On the one side
there is data mining as synonym for KDD, meaning that data mining contains all aspects
of the knowledge discovery process. This definition is in particular common in practice
and frequently leads to problems to distinguish the terms clearly. The second way
of looking at it considers data mining as part of the KDD-Processes (see [FPSS96])
and describes the modelling phase, i.e. the application of algorithms and methods for
the calculation of the searched patterns or models. Other authors like for instance
Kumar and Joshi [KJ03] consider data mining in addition as the search for valuable
information in large quantities of data. In this article, we equate data mining with the
modelling phase of the KDD process.
The roots of data mining lie in most diverse areas of research, which underlines the
interdisciplinary character of this field. In the following we briefly discuss the relations
to three of the addressed research areas: Databases, machine learning and statistics.
Databases are necessary in order to analyze large quantities of data efficiently. In

3
this connection, a database represents not only the medium for consistent storing and
accessing, but moves in the closer interest of research, since the analysis of the data
with data mining algorithms can be supported by databases and thus the use of database
technology in the data mining process might be useful. An overview of data mining
from the database perspective can be found in [CHY96].
Machine Learning (ML) is an area of artificial intelligence concerned with the de-
velopment of techniques which allow computers to ”learn” by the analysis of data sets.
The focus of most machine learning methods is on symbolic data. ML is also con-
cerned with the algorithmic complexity of computational implementations. Mitchell
presents many of the commonly used ML methods in [Mit97].
Statistics has its grounds in mathematics and deals with the science and practice for
the analysis of empirical data. It is based on statistical theory which is a branch of ap-
plied mathematics. Within statistical theory, randomness and uncertainty are modelled
by probability theory. Today many methods of statistics are used in the field of KDD.
Good overviews are given in [HTF01, Be99, Mai02].

1.3 Definition of Text Mining

Text mining or knowledge discovery from text (KDT) — for the first time mentioned
in Feldman et al. [FD95] — deals with the machine supported analysis of text. It uses
techniques from information retrieval, information extraction as well as natural lan-
guage processing (NLP) and connects them with the algorithms and methods of KDD,
data mining, machine learning and statistics. Thus, one selects a similar procedure as
with the KDD process, whereby not data in general, but text documents are in focus
of the analysis. From this, new questions for the used data mining methods arise. One
problem is that we now have to deal with problems of — from the data modelling
perspective — unstructured data sets.
If we try to define text mining, we can refer to related research areas. For each
of them, we can give a different definition of text mining, which is motivated by the
specific perspective of the area:

Text Mining = Information Extraction. The first approach assumes that text mining
essentially corresponds to information extraction (cf. section 3.3) — the extrac-
tion of facts from texts.
Text Mining = Text Data Mining. Text mining can be also defined — similar to data
mining — as the application of algorithms and methods from the fields machine
learning and statistics to texts with the goal of finding useful patterns. For this
purpose it is necessary to pre-process the texts accordingly. Many authors use
information extraction methods, natural language processing or some simple pre-
processing steps in order to extract data from texts. To the extracted data then
data mining algorithms can be applied (see [NM02, Gai03]).
Text Mining = KDD Process. Following the knowledge discovery process model [cri99],
we frequently find in literature text mining as a process with a series of partial
steps, among other things also information extraction as well as the use of data
mining or statistical procedures. Hearst summarizes this in [Hea99] in a general

4
 manner as the extraction of not yet discovered information in large collections of
texts. Also Kodratoff in [Kod99] and Gomez in [Hid02] consider text mining as
process orientated approach on texts.
In this article, we consider text mining mainly as text data mining. Thus, our focus
is on methods that extract useful patterns from texts in order to, e.g., categorize or
structure text collections or to extract useful information.

1.4 Related Research Areas

Current research in the area of text mining tackles problems of text representation,
classification, clustering, information extraction or the search for and modelling of
hidden patterns. In this context the selection of characteristics and also the influence of
domain knowledge and domain-specific procedures plays an important role. Therefore,
an adaptation of the known data mining algorithms to text data is usually necessary. In
order to achieve this, one frequently relies on the experience and results of research in
information retrieval, natural language processing and information extraction. In all of
these areas we also apply data mining methods and statistics to handle their specific
tasks:

Information Retrieval (IR). Information retrieval is the finding of documents which

contain answers to questions and not the finding of answers itself [Hea99]. In order to
achieve this goal statistical measures and methods are used for the automatic process-
ing of text data and comparison to the given question. Information retrieval in the
broader sense deals with the entire range of information processing, from data retrieval
to knowledge retrieval (see [SJW97] for an overview). Although, information retrieval
is a relatively old research area where first attempts for automatic indexing where made
in 1975 [SWY75], it gained increased attention with the rise of the World Wide Web
and the need for sophisticated search engines.
Even though, the definition of information retrieval is based on the idea of ques-
tions and answers, systems that retrieve documents based on keywords, i.e. systems
that perform document retrieval like most search engines, are frequently also called
information retrieval systems.

Natural Language Processing (NLP). The general goal of NLP is to achieve a better
understanding of natural language by use of computers [Kod99]. Others include also
the employment of simple and durable techniques for the fast processing of text, as
they are presented e.g. in [Abn91]. The range of the assigned techniques reaches from
the simple manipulation of strings to the automatic processing of natural language
inquiries. In addition, linguistic analysis techniques are used among other things for
the processing of text.

Information Extraction (IE). The goal of information extraction methods is the ex-
traction of specific information from text documents. These are stored in data base-like
patterns (see [Wil97]) and are then available for further use. For further details see
section 3.3.

5
 In the following, we will frequently refer to the above mentioned related areas of
research. We will especially provide examples for the use of machine learning methods
in information extraction and information retrieval.

2 Text Encoding
For mining large document collections it is necessary to pre-process the text documents
and store the information in a data structure, which is more appropriate for further pro-
cessing than a plain text file. Even though, meanwhile several methods exist that try to
exploit also the syntactic structure and semantics of text, most text mining approaches
are based on the idea that a text document can be represented by a set of words, i.e.
a text document is described based on the set of words contained in it (bag-of-words
representation). However, in order to be able to define at least the importance of a word
within a given document, usually a vector representation is used, where for each word a
numerical ”importance” value is stored. The currently predominant approaches based
on this idea are the vector space model [SWY75], the probabilistic model [Rob77] and
the logical model [van86].
In the following we briefly describe, how a bag-of-words representation can be
obtained. Furthermore, we describe the vector space model and corresponding sim-
ilarity measures in more detail, since this model will be used by several text mining
approaches discussed in this article.

2.1 Text Preprocessing

In order to obtain all words that are used in a given text, a tokenization process is re-
quired, i.e. a text document is split into a stream of words by removing all punctuation
marks and by replacing tabs and other non-text characters by single white spaces. This
tokenized representation is then used for further processing. The set of different words
obtained by merging all text documents of a collection is called the dictionary of a
document collection.
In order to allow a more formal description of the algorithms, we define first some
terms and variables that will be frequently used in the following: Let D be the set of
documents and T = {t1 , . . . , tm } be the dictionary, i.e. the set of all different terms
occurring in D, then the absolute frequency of term t ∈ T in document d ∈ D is given
by tf(d, t). We denote the term vectors t~d = (tf(d, t1 ), . . . , tf(d, tm )). Later on, we will
P of the centroid of a set X of term vectors. It is defined as the mean
also need the notion
value t~X := |X| 1 ~
t~d ∈X td of its term vectors. In the sequel, we will apply tf also on
P
subsets of terms: For T ⊆ T, we let tf(d, T0 ) := t∈T0 tf(d, t).
0

2.1.1 Filtering, Lemmatization and Stemming

In order to reduce the size of the dictionary and thus the dimensionality of the descrip-
tion of documents within the collection, the set of words describing the documents can
be reduced by filtering and lemmatization or stemming methods.

6
 Filtering methods remove words from the dictionary and thus from the documents.
A standard filtering method is stop word filtering. The idea of stop word filtering is
to remove words that bear little or no content information, like articles, conjunctions,
prepositions, etc. Furthermore, words that occur extremely often can be said to be of
little information content to distinguish between documents, and also words that occur
very seldom are likely to be of no particular statistical relevance and can be removed
from the dictionary [FBY92]. In order to further reduce the number of words in the
dictionary, also (index) term selection methods can be used (see Sect. 2.1.2).
Lemmatization methods try to map verb forms to the infinite tense and nouns to
the singular form. However, in order to achieve this, the word form has to be known,
i.e. the part of speech of every word in the text document has to be assigned. Since
this tagging process is usually quite time consuming and still error-prone, in practice
frequently stemming methods are applied.
Stemming methods try to build the basic forms of words, i.e. strip the plural ’s’ from
nouns, the ’ing’ from verbs, or other affixes. A stem is a natural group of words with
equal (or very similar) meaning. After the stemming process, every word is represented
by its stem. A well-known rule based stemming algorithm has been originally proposed
by Porter [Por80]. He defined a set of production rules to iteratively transform (English)
words into their stems.

2.1.2 Index Term Selection

To further decrease the number of words that should be used also indexing or keyword
selection algorithms can be used (see, e.g. [DDFL90, WMB99]). In this case, only the
selected keywords are used to describe the documents. A simple method for keyword
selection is to extract keywords based on their entropy. E.g. for each word t in the
vocabulary the entropy as defined by [LS89] can be computed:

1 X tf(d, t)
W (t) = 1 + P (d, t) log2 P (d, t) with P (d, t) = Pn (1)
log2 |D| l=1 tf(dl , t)
d∈D

Here the entropy gives a measure how well a word is suited to separate documents
by keyword search. For instance, words that occur in many documents will have low
entropy. The entropy can be seen as a measure of the importance of a word in the given
domain context. As index words a number of words that have a high entropy relative to
their overall frequency can be chosen, i.e. of words occurring equally often those with
the higher entropy can be preferred.
In order to obtain a fixed number of index terms that appropriately cover the docu-
ments, a simple greedy strategy can be applied: From the first document in the collec-
tion select the term with the highest relative entropy (or information gain as described
in Sect. 3.1.1) as an index term. Then mark this document and all other documents con-
taining this term. From the first of the remaining unmarked documents select again the
term with the highest relative entropy as an index term. Then mark again this document
and all other documents containing this term. Repeat this process until all documents
are marked, then unmark them all and start again. The process can be terminated when
the desired number of index terms have been selected. A more detailed discussion of

7
the benefits of this approach for clustering - with respect to reduction of words required
in order to obtain a good clustering performance - can be found in [BN04].
An index term selection methods that is more appropriate if we have to learn a
classifier for documents is discussed in Sect. 3.1.1. This approach also considers the
word distributions within the classes.

2.2 The Vector Space Model

Despite of its simple data structure without using any explicit semantic information,
the vector space model enables very efficient analysis of huge document collections. It
was originally introduced for indexing and information retrieval [SWY75] but is now
used also in several text mining approaches as well as in most of the currently available
document retrieval systems.
The vector space model represents documents as vectors in m-dimensional space,
i.e. each document d is described by a numerical feature vector w(d) = (x(d, t1 ), . . . , x(d, tm )).
Thus, documents can be compared by use of simple vector operations and even queries
can be performed by encoding the query terms similar to the documents in a query
vector. The query vector can then be compared to each document and a result list can
be obtained by ordering the documents according to the computed similarity [SAB94].
The main task of the vector space representation of documents is to find an appropriate
encoding of the feature vector.
Each element of the vector usually represents a word (or a group of words) of the
document collection, i.e. the size of the vector is defined by the number of words (or
groups of words) of the complete document collection. The simplest way of document
encoding is to use binary term vectors, i.e. a vector element is set to one if the corre-
sponding word is used in the document and to zero if the word is not. This encoding
will result in a simple Boolean comparison or search if a query is encoded in a vector.
Using Boolean encoding the importance of all terms for a specific query or comparison
is considered as similar. To improve the performance usually term weighting schemes
are used, where the weights reflect the importance of a word in a specific document of
the considered collection. Large weights are assigned to terms that are used frequently
in relevant documents but rarely in the whole document collection [SB88]. Thus a
weight w(d, t) for a term t in document d is computed by term frequency tf(d, t) times
inverse document frequency idf(t), which describes the term specificity within the doc-
ument collection. In [SAB94] a weighting scheme was proposed that has meanwhile
proven its usability in practice. Besides term frequency and inverse document fre-
quency — defined as idf (t) := log(N/nt ) —, a length normalization factor is used to
ensure that all documents have equal chances of being retrieved independent of their
lengths:

tf(d, t) log(N/nt )
w(d, t) = qP , (2)
m 2 2
j=1 tf (d, tj ) (log(N/ntj ))

where N is the size of the document collection D and nt is the number of documents
in D that contain term t.

8
 Based on a weighting scheme a document d is defined by a vector of term weights
w(d) = (w(d, t1 ), . . . , w(d, tm )) and the similarity S of two documents d1 and d2
(or the similarity of a document and a query vector) can be computed based on the
inner product of the vectors (by which – if we assume normalized vectors – the cosine
between the two document vectors is computed), i.e.
Xm
S(d1 , d2 ) = w(d1 , tk ) · w(d2 , tk ). (3)
k=1

A frequently used distance measure is the Euclidian distance. We calculate the

distance between two text documents d1 , d2 ∈ D as follows:
r
Xm 2
dist(d1 , d2 ) = 2 |w(d1 , tk ) − w(d2 , tk )| . (4)
k=1

However, the Euclidean distance should only be used for normalized vectors, since
otherwise the different lengths of documents can result in a smaller distance between
documents that share less words than between documents that have more words in
common and should be considered therefore as more similar.
Note that for normalized vectors the scalar product is not much different in behavior
from the Euclidean distance, since for two vectors ~x and ~y it is
µ ¶
~x~y 1 ~x ~y
cos ϕ = = 1 − d2 , .
|~x| · |~y | 2 |~x| |~y |
For a more detailed discussion of the vector space model and weighting schemes
see, e.g. [BYRN99, Gre98, SB88, SWY75].

2.3 Linguistic Preprocessing

Often text mining methods may be applied without further preprocessing. Sometimes,
however, additional linguistic preprocessing (c.f. [MS01a]) may be used to enhance the
available information about terms. For this, the following approaches are frequently
applied:
Part-of-speech tagging (POS) determines the part of speech tag, e.g. noun, verb,
adjective, etc. for each term.
Text chunking aims at grouping adjacent words in a sentence. An example of a chunk
is the noun phrase “the current account deficit”.
Word Sense Disambiguation (WSD) tries to resolve the ambiguity in the meaning of
single words or phrases. An example is ‘bank’ which may have – among others –
the senses ‘financial institution’ or the ‘border of a river or lake’. Thus, instead of
terms the specific meanings could be stored in the vector space representation.
This leads to a bigger dictionary but considers the semantic of a term in the
representation.
Parsing produces a full parse tree of a sentence. From the parse, we can find the
relation of each word in the sentence to all the others, and typically also its
function in the sentence (e.g. subject, object, etc.).

9
 Linguistic processing either uses lexica and other resources as well as hand-crafted
rules. If a set of examples is available machine learning methods as described in section
3, especially in section 3.3, may be employed to learn the desired tags.
It turned out, however, that for many text mining tasks linguistic preprocessing is of
limited value compared to the simple bag-of-words approach with basic preprocessing.
The reason is that the co-occurrence of terms in the vector representation serves as
an automatic disambiguation, e.g. for classification [LK02]. Recently some progress
was made by enhancing bag of words with linguistic feature for text clustering and
classification [HSS03, BH04].

3 Data Mining Methods for Text

One main reason for applying data mining methods to text document collections is to
structure them. A structure can significantly simplify the access to a document collec-
tion for a user. Well known access structures are library catalogues or book indexes.
However, the problem of manual designed indexes is the time required to maintain
them. Therefore, they are very often not up-to-date and thus not usable for recent pub-
lications or frequently changing information sources like the World Wide Web. The
existing methods for structuring collections either try to assign keywords to documents
based on a given keyword set (classification or categorization methods) or automat-
ically structure document collections to find groups of similar documents (clustering
methods). In the following we first describe both of these approaches. Furthermore,
we discuss in Sect. 3.3 methods to automatically extract useful information patterns
from text document collections. In Sect. 3.4 we review methods for visual text min-
ing. These methods allow in combination with structuring methods the development
of powerful tools for the interactive exploration of document collections. We conclude
this section with a brief discussion of further application areas for text mining.

3.1 Classification
Text classification aims at assigning pre-defined classes to text documents [Mit97]. An
example would be to automatically label each incoming news story with a topic like
”sports”, ”politics”, or ”art”. Whatever the specific method employed, a data mining
classification task starts with a training set D = (d1 , . . . , dn ) of documents that are
already labelled with a class L ∈ L (e.g. sport, politics). The task is then to determine
a classification model
f :D→L f (d) = L (5)
which is able to assign the correct class to a new document d of the domain.
To measure the performance of a classification model a random fraction of the la-
belled documents is set aside and not used for training. We may classify the documents
of this test set with the classification model and compare the estimated labels with
the true labels. The fraction of correctly classified documents in relation to the total
number of documents is called accuracy and is a first performance measure.
Often, however, the target class covers only a small percentage of the documents.
Then we get a high accuracy if we assign each document to the alternative class. To

10
avoid this effect different measures of classification success are often used. Precision
quantifies the fraction of retrieved documents that are in fact relevant, i.e. belong to the
target class. Recall indicates which fraction of the relevant documents is retrieved.

#{relevant ∩ retrieved} #{relevant ∩ retrieved}

precision = recall = (6)
#retrieved #relevant
Obviously there is a trade off between precision and recall. Most classifiers inter-
nally determine some “degree of membership” in the target class. If only documents of
high degree are assigned to the target class, the precision is high. However, many rele-
vant documents might have been overlooked, which corresponds to a low recall. When
on the other hand the search is more exhaustive, recall increases and precision goes
down. The F-score is a compromise of both for measuring the overall performance of
classifiers.
2
F = (7)
1/recall + 1/precision

3.1.1 Index Term Selection

As document collections often contain more than 100000 different words we may select
the most informative ones for a specific classification task to reduce the number of
words and thus the complexity of the classification problem at hand. One commonly
used ranking score is the information gain which for a term tj is defined as
2
X X1 X2
1 1
IG(tj ) = p(Lc ) log2 − p(tj =m) p(Lc |tj =m) log2
c=1
p(L c ) m=0 c=1
p(L |t
c j =m)
(8)
Here p(Lc ) is the fraction of training documents with classes L1 and L2 , p(tj =1) and
p(tj =0) is the number of documents with / without term tj and p(Lc |tj =m) is the
conditional probability of classes L1 and L2 if term tj is contained in the document or
is missing. It measures how useful tj is for predicting L1 from an information-theoretic
point of view. We may determine IG(tj ) for all terms and remove those with very low
information gain from the dictionary.
In the following sections we describe the most frequently used data mining methods
for text categorization.

3.1.2 Naı̈ve Bayes Classifier

Probabilistic classifiers start with the assumption that the words of a document di have
been generated by a probabilistic mechanism. It is supposed that the class L(di ) of
document di has some relation to the words which appear in the document. This may
be described by the conditional distribution p(t1 , . . . , tni |L(di )) of the ni words given
the class. Then the Bayesian formula yields the probability of a class given the words
of a document [Mit97]

p(t1 , . . . , tni |Lc )p(Lc )

p(Lc |t1 , . . . , tni ) = P
L∈L p(t1 , . . . , tni |L)p(L)

11
Note that each document is assumed to belong to exactly one of the k classes in L.
The prior probability p(L) denotes the probability that an arbitrary document belongs
to class L before its words are known. Often the prior probabilities of all classes may
be taken to be equal. The conditional probability on the left is the desired posterior
probability that the document with words t1 , . . . , tni belongs to class Lc . We may
assign the class with highest posterior probability to our document.
For document classification it turned out that the specific order of the words in a
document is not very important. Even more we may assume that for documents of a
given class a word appears in the document irrespective of the presence of other words.
This leads to a simple formula for the conditional probability of words given a class Lc
ni
Y
p(t1 , . . . , tni |Lc ) = p(tj |Lc )
j=1

Combining this “naı̈ve” independence assumption with the Bayes formula defines the
Naı̈ve Bayes classifier [Goo65]. Simplifications of this sort are required as many thou-
sand different words occur in a corpus.
The naı̈ve Bayes classifier involves a learning step which simply requires the esti-
mation of the probabilities of words p(tj |Lc ) in each class by its relative frequencies
in the documents of a training set which are labelled with Lc . In the classification step
the estimated probabilities are used to classify a new instance according to the Bayes
rule. In order to reduce the number of probabilities p(tj |Lm ) to be estimated, we can
use index term selection methods as discussed above in Sect. 3.1.1.
Although this model is unrealistic due to its restrictive independence assumption
it yields surprisingly good classifications [DPHS98, Joa98]. It may be extended into
several directions [Seb02].
As the effort for manually labeling the documents of the training set is high, some
authors use unlabeled documents for training. Assume that from a small training set
it has been established that word ti is highly correlated with class Lc . If from unla-
beled documents it may be determined that word tj is highly correlated with ti , then
also tj is a good predictor for class Lc . In this way unlabeled documents may im-
prove classification performance. In [NMTM00] the authors used a combination of
Expectation-Maximization (EM) [DLR77] and a naı̈ve Bayes classifier and were able
to reduce the classification error by up to 30%.

3.1.3 Nearest Neighbor Classifier

Instead of building explicit models for the different classes we may select documents
from the training set which are “similar” to the target document. The class of the
target document subsequently may be inferred from the class labels of these similar
documents. If k similar documents are considered, the approach is also known as k-
nearest neighbor classification.
There is a large number of similarity measures used in text mining. One possibility
is simply to count the number of common words in two documents. Obviously this
has to be normalized to account for documents of different lengths. On the other hand
words have greatly varying information content. A standard way to measure the latter

12
is the cosine similarity as defined in (3). Note that only a small fraction of all possible
terms appear in this sums as w(d, t) = 0 if the term t is not present in the document d.
Other similarity measures are discussed in [BYRN99].
For deciding whether document di belongs to class Lm , the similarity S(di , dj )
to all documents dj in the training set is determined. The k most similar training
documents (neighbors) are selected. The proportion of neighbors having the same
class may be taken as an estimator for the probability of that class, and the class with
the largest proportion is assigned to document di . The optimal number k of neighbors
may be estimated from additional training data by cross-validation.
Nearest neighbor classification is a nonparametric method and it can be shown that
for large data sets the error rate of the 1-nearest neighbor classifier is never larger
than twice the optimal error rate [HTF01]. Several studies have shown that k-nearest
neighbor methods have very good performance in practice [Joa98]. Their drawback
is the computational effort during classification, where basically the similarity of a
document with respect to all other documents of a training set has to be determined.
Some extensions are discussed in [Seb02].

3.1.4 Decision Trees

Decision trees are classifiers which consist of a set of rules which are applied in a
sequential way and finally yield a decision. They can be best explained by observing
the training process, which starts with a comprehensive training set. It uses a divide and
conquer strategy: For a training set M with labelled documents the word ti is selected,
which can predict the class of the documents in the best way, e.g. by the information
gain (8). Then M is partitioned into two subsets, the subset Mi+ with the documents
containing ti , and the subset Mi− with the documents without ti . This procedure is
recursively applied to Mi+ and Mi− . It stops if all documents in a subset belong to the
same class Lc . It generates a tree of rules with an assignment to actual classes in the
leaves.
Decision trees are a standard tool in data mining [Qui86, Mit97]. They are fast and
scalable both in the number of variables and the size of the training set. For text mining,
however, they have the drawback that the final decision depends only on relatively few
terms. A decisive improvement may be achieved by boosting decision trees [SS99],
i.e. determining a set of complementary decision trees constructed in such a way that
the overall error is reduced. [SS00] use even simpler one step decision trees containing
only one rule and get impressive results for text classification.

3.1.5 Support Vector Machines and Kernel Methods

A Support Vector Machine (SVM) is a supervised classification algorithm that recently
has been applied successfully to text classification tasks [Joa98, DPHS98, LK02]. As
usual a document d is represented by a – possibly weighted – vector (td1 , . . . , tdN ) of
the counts of its words. A single SVM can only separate two classes — a positive class
L1 (indicated by y = +1) and a negative class L2 (indicated by y = −1). In the space
of input vectors a hyperplane may be defined by setting y = 0 in the following linear

13
 documents of class 1

h yp
erp
lan
e

ma
rg in x

ma
rg in x

documents of class 2

Figure 2: Hyperplane with maximal distance (margin) to examples of positive and

negative classes constructed by the support vector machine.

equation.
N
X
y = f (t~d ) = b0 + bj tdj
j=1

The SVM algorithm determines a hyperplane which is located between the positive and
negative examples of the training set. The parameters bj are adapted in such a way that
the distance ξ – called margin – between the hyperplane and the closest positive and
negative example documents is maximized, as shown in Fig. 3.1.5. This amounts to a
constrained quadratic optimization problem which can be solved efficiently for a large
number of input vectors.
The documents having distance ξ from the hyperplane are called support vectors
and determine the actual location of the hyperplane. Usually only a small fraction of
documents are support vectors. A new document with term vector t~d is classified in
L1 if the value f (t~d ) > 0 and into L2 otherwise. In case that the document vectors of
the two classes are not linearly separable a hyperplane is selected such that as few as
possible document vectors are located on the “wrong” side.
SVMs can be used with non-linear predictors by transforming the usual input fea-
tures in a non-linear way, e.g. by defining a feature map
¡ ¢
φ(t1 , . . . , tN ) = t1 , . . . , tN , t21 , t1 t2 , . . . , tN tN −1 , t2N

Subsequently a hyperplane may be defined in the expanded input space. Obviously

such non-linear transformations may be defined in a large number of ways.
The most important property of SVMs is that learning is nearly independent of the
dimensionality of the feature space. It rarely requires feature selection as it inherently
selects data points (the support vectors) required for a good classification. This allows
good generalization even in the presence of a large number of features and makes SVM

14
especially suitable for the classification of texts [Joa98]. In the case of textual data the
choice of the kernel function has a minimal effect on the accuracy of classification:
Kernels that imply a high dimensional feature space show slightly better results in
terms of precision and recall, but they are subject to overfitting [LK02].

3.1.6 Classifier Evaluations

During the last years text classifiers have been evaluated on a number of benchmark
document collections. It turns out that the level of performance of course depends
on the document collection. Table 1 gives some representative results achieved for
the Reuters 20 newsgroups collection [Seb02, p.38]. Concerning the relative quality
of classifiers boosted trees, SVMs, and k-nearest neighbors usually deliver top-notch
performance, while naı̈ve Bayes and decision trees are less reliable.

Table 1: Performance of Different Classifiers for the Reuters collection

Method F1 -value
naı̈ve Bayes 0.795
decision tree C4.5 0.794
k-nearest neighbor 0.856
SVM 0.870
boosted tree 0.878

3.2 Clustering
Clustering method can be used in order to find groups of documents with similar con-
tent. The result of clustering is typically a partition (also called) clustering P, a set
of clusters P . Each cluster consists of a number of documents d. Objects — in our
case documents — of a cluster should be similar and dissimilar to documents of other
clusters. Usually the quality of clusterings is considered better if the contents of the
documents within one cluster are more similar and between the clusters more dissimi-
lar. Clustering methods group the documents only by considering their distribution in
document space (for example, a n-dimensional space if we use the vector space model
for text documents).
Clustering algorithms compute the clusters based on the attributes of the data and
measures of (dis)similarity. However, the idea of what an ideal clustering result should
look like varies between applications and might be even different between users. One
can exert influence on the results of a clustering algorithm by using only subsets of
attributes or by adapting the used similarity measures and thus control the clustering
process. To which extent the result of the cluster algorithm coincides with the ideas
of the user can be assessed by evaluation measures. A survey of different kinds of
clustering algorithms and the resulting cluster types can be found in [SEK03].
In the following, we first introduce standard evaluation methods and present then
details for hierarchical clustering approaches, k-means, bi-section-k-means, self-organizing

15
maps and the EM-algorithm. We will finish the clustering section with a short overview
of other clustering approaches used for text clustering.

3.2.1 Evaluation of clustering results

In general, there are two ways to evaluate clustering results. One the one hand statistical
measures can be used to describe the properties of a clustering result. On the other hand
some given classification can be seen as a kind of gold standard which is then typically
used to compare the clustering results with the given classification. We discuss both
aspects in the following.

Statistical Measures In the following, we first discuss measures which cannot make
use of a given classification L of the documents. They are called indices in statistical
literature and evaluate the quality of a clustering on the basis of statistic connections.
One finds a large number of indices in literature (see [Fic97, DH73]). One of the
most well-known measures is the mean square error. It permits to make statements
on quality of the found clusters dependent on the number of clusters. Unfortunately,
the computed quality is always better if the number of cluster is higher. In [KR90] an
alternative measure, the silhouette coefficient, is presented which is independent of the
number of clusters. We introduce both measures in the following.

Mean square error If one keeps the number of dimensions and the number of clus-
ters constant the mean square error (Mean Square error, MSE) can be used likewise for
the evaluation of the quality of clustering. The mean square error is a measure for the
compactness of the clustering and is defined as follows:

Definition 1 (MSE) The means square error (M SE) for a given clustering P is de-
fined as X
M SE(P) = M SE(P ), (9)
P ∈P

whereas the means square error for a cluster P is given by:

X
M SE(P ) = dist(d, µP )2 , (10)
d∈P

P
and µP = 1 ~
d∈P td is the centroid of the clusters P and dist is a distance measure.
|P |

Silhouette Coefficient One clustering measure that is independent from the number
of clusters is the silhouette coefficient SC(P) (cf. [KR90]). The main idea of the coef-
ficient is to find out the location of a document in the space with respect to the cluster
of the document and the next similar cluster. For a good clustering the considered doc-
ument is nearby the own cluster whereas for a bad clustering the document is closer
to the next cluster. With the help of the silhouette coefficient one is able to judge the
quality of a cluster or the entire clustering (details can be found in [KR90]). [KR90]
gives characteristic values of the silhouette coefficient for the evaluation of the cluster

16
quality. A value for SC(P) between 0.7 and 1.0 signals excellent separation between
the found clusters, i.e. the objects within a cluster are very close to each other and
are far away from other clusters. The structure was very well identified by the cluster
algorithm. For the range from 0.5 to 0.7 the objects are clearly assigned to the appro-
priate clusters. A larger level of noise exists in the data set if the silhouette coefficient
is within the range of 0.25 to 0.5 whereby also here still clusters are identifiable. Many
objects could not be assigned clearly to one cluster in this case due to the cluster algo-
rithm. At values under 0.25 it is practically impossible to identify a cluster structure
and to calculate meaningful (from the view of application) cluster centers. The cluster
algorithm more or less ”guessed” the clustering.

Comparative Measures The purity measure is based on the well-known precision

measure for information retrieval (cf. [PL02]). Each resulting cluster P from a parti-
tioning P of the overall document set D is treated as if it were the result of a query.
Each set L of documents of a partitioning L, which is obtained by manual labelling,
is treated as if it is the desired set of documents for a query which leads to the same
definitions for precision, recall and f-score as defined in Equations 6 and 7. The two
partitions P and L are then compared as follows.
The precision of a cluster P ∈ P for a given category L ∈ L is given by
|P ∩ L|
Precision(P, L) := . (11)
|P |
The overall value for purity is computed by taking the weighted average of maximal
precision values:
X |P |
Purity(P, L) := max Precision(P, L). (12)
|D| L∈L
P ∈P

The counterpart of purity is:

X |L|
InversePurity(P, L) := max Recall(P, L), (13)
|D| P ∈P
L∈L

where Recall(P, L) := Precision(L, P ) and the well known

X |L| 2 · Recall(P, L) · Precision(P, L)
F-Measure(P, L) := max , (14)
|D| P ∈P Recall(P, L) + Precision(P, L)
L∈L

which is based on the F-score as defined in Eq. 7.

The three measures return values in the interval [0, 1], with 1 indicating optimal
agreement. Purity measures the homogeneity of the resulting clusters when evaluated
against a pre-categorization, while inverse purity measures how stable the pre-defined
categories are when split up into clusters. Thus, purity achieves an “optimal” value
of 1 when the number of clusters k equals |D|, whereas inverse purity achieves an
“optimal” value of 1 when k equals 1. Another name in the literature for inverse purity
is microaveraged precision. The reader may note that, in the evaluation of clustering

17
results, microaveraged precision is identical to microaveraged recall (cf. e.g. [Seb02]).
The F-measure works similar as inverse purity, but it depreciates overly large clusters,
as it includes the individual precision of these clusters into the evaluation.
While (inverse) purity and F-measure only consider ‘best’ matches between ‘queries’
and manually defined categories, the entropy indicates how large the information con-
tent uncertainty of a clustering result with respect to the given classification is
X
E(P, L) = prob(P ) · E(P ), where (15)
P ∈P
X
E(P ) = − prob(L|P ) log(prob(L|P )) (16)
L∈L
|P |
where prob(L|P ) = Precision(P, L) and prob(P ) = |D| . The entropy has the range
[0, log(|L|)], with 0 indicating optimality.

3.2.2 Partitional Clustering

Hierarchical Clustering Algorithms [MS01a, SKK00] got their name since they
form a sequence of groupings or clusters that can be represented in a hierarchy of clus-
ters. This hierarchy can be obtained either in a top-down or bottom-up fashion. Top-
down means that we start with one cluster that contains all documents. This cluster
is stepwise refined by splitting it iteratively into sub-clusters. One speaks in this case
also of the so called ”divisive” algorithm. The bottom-up or ”agglomerative” proce-
dures start by considering every document as individual cluster. Then the most similar
clusters are iteratively merged, until all documents are contained in one single cluster.
In practice the divisive procedure is almost of no importance due to its generally bad
results. Therefore, only the agglomerative algorithm is outlined in the following.
The agglomerative procedure considers initially each document d of the the whole
document set D as an individual cluster. It is the first cluster solution. It is assumed
that each document is member of exactly one cluster. One determines the similarity
between the clusters on the basis of this first clustering and selects the two clusters p,
q of the clustering P with the minimum distance dist(p, q). Both cluster are merged
and one receives a new clustering. One continues this procedure and re-calculates the
distances between the new clusters in order to join again the two clusters with the
minimum distance dist(p, q). The algorithm stops if only one cluster is remaining.
The distance can be computed according to Eq. 4. It is also possible to derive
the clusters directly on the basis of the similarity relationship given by a matrix. For
the computation of the similarity between clusters that contain more than one element
different distance measures for clusters can be used, e.g. based on the outer cluster
shape or the cluster center. Common linkage procedures that make use of different
cluster distance measures are single linkage, average linkage or Ward’s procedure. The
obtained clustering depends on the used measure. Details can be found, for example,
in [DH73].
By means of so-called dendrograms one can represent the hierarchy of the clusters
obtained as a result of the repeated merging of clusters as described above. The dendro-
grams allows to estimate the number of clusters based on the distances of the merged

18
clusters. Unfortunately, the selection of the appropriate linkage method depends on the
desired cluster structure, which is usually unknown in advance. For example, single
linkage tends to follow chain-like clusters in the data, while complete linkage tends
to create ellipsoid clusters. Thus prior knowledge about the expected distribution and
cluster form is usually necessary for the selection of the appropriate method (see also
[DH73]). However, substantially more problematic for the use of the algorithm for
large data sets is the memory required to store the similarity matrix, which consists of
n(n − 1)/2 elements where n is the number of documents. Also the runtime behavior
with O(n2 ) is worse compared to the linear behavior of k-means as discussed in the
following.

k-means is one of the most frequently used clustering algorithms in practice in the
field of data mining and statistics (see [DH73, Har75]). The procedure which originally
comes from statistics is simple to implement and can also be applied to large data sets.
It turned out that especially in the field of text clustering k-means obtains good results.
Proceeding from a starting solution in which all documents are distributed on a given
number of clusters one tries to improve the solution by a specific change of the alloca-
tion of documents to the clusters. Meanwhile, a set of variants exists whereas the basic
principle goes back to Forgy 1965 [For65] or MacQueen 1967 [Mac67]. In literature
for vector quantization k-means is also known under the name LloydMaxAlgorithm
([GG92]). The basic principle is shown in the following algorithm:

Algorithm 1 The k-means algorithm

Input: set D, distance measure dist, number k of cluster
Output: ASpartitioning P of the set D of documents (i. e., a set P of k disjoint subsets
of D with P ∈P P = D).
1: Choose randomly k data points from D as starting centroids t~ P1 . . . t~
Pk .
2: repeat
3: Assign each point of P to the closest centroid with respect to dist.
4: (Re-)calculate the cluster centroids t~P1 . . . t~Pk of clusters P1 . . . Pk .
5: until cluster centroids t~P1 . . . t~
Pk are stable
6: return set P := {P1 , . . . , Pk }, of clusters.

k-means essentially consists of the steps three and four in the algorithm, whereby
the number of clusters k must be given. In step three the documents are assigned
to the nearest of the k centroids (also called cluster prototype). Step four calculates
a new centroids on the basis of the new allocations. We repeat the two steps in a
loop (step five) until the cluster centroids do not change any more. The algorithm 5.1
corresponds to a simple hill climbing procedure which typically gets stuck in a local
optimum (the finding of the global optimum is a NP complete problem). Apart from
a suitable method to determine the starting solution (step one), we require a measure
for calculating the distance or similarity in step three (cf. section 2.1). Furthermore the
abort criterion of the loop in step five can be chosen differently e.g. by stopping after a
fix number of iterations.

19
Bi-Section-k-means One fast text clustering algorithm, which is also able to deal
with the large size of the textual data is the Bi-Section-k-means algorithm. In [SKK00]
it was shown that Bi-Section-k-means is a fast and high-quality clustering algorithm
for text documents which is frequently outperforming standard k-means as well as
agglomerative clustering techniques.
Bi-Section-k-means is based on the k-means algorithm. It repeatedly splits the
largest cluster (using k-means) until the desired number of clusters is obtained. Another
way of choosing the next cluster to be split is picking the one with the largest variance.
[SKK00] showed neither of these two has a significant advantage.

Self Organizing Map (SOM) [Koh82] are a special architecture of neural networks
that cluster high-dimensional data vectors according to a similarity measure. The clus-
ters are arranged in a low-dimensional topology that preserves the neighborhood re-
lations in the high dimensional data. Thus, not only objects that are assigned to one
cluster are similar to each other (as in every cluster analysis), but also objects of nearby
clusters are expected to be more similar than objects in more distant clusters. Usually,
two-dimensional grids of squares or hexagons are used (cf. Fig. 3).
The network structure of a self-organizing map has two layers (see Fig. 3). The
neurons in the input layer correspond to the input dimensions, here the words of the
document vector. The output layer (map) contains as many neurons as clusters needed.
All neurons in the input layer are connected with all neurons in the output layer. The
weights of the connection between input and output layer of the neural network encode
positions in the high-dimensional data space (similar to the cluster prototypes in k-
means). Thus, every unit in the output layer represents a cluster center. Before the
learning phase of the network, the two-dimensional structure of the output units is fixed
and the weights are initialized randomly. During learning, the sample vectors (defining
the documents) are repeatedly propagated through the network. The weights of the
most similar prototype w~s (winner neuron) are modified such that the prototype moves
toward the input vector w ~i , which is defined by the currently considered document
~i := t~d (competitive learning). As similarity measure usually the Euclidean
d, i.e. w
distance is used. However, for text documents the scalar product (see Eq. 3) can be
applied. The weights w ~ s of the winner neuron are modified according to the following
equation:
w~s 0 = w~s + σ · (w~s − w
~i ),
where σ is a learning rate.
To preserve the neighborhood relations, prototypes that are close to the winner
neuron in the two-dimensional structure are also moved in the same direction. The
weight change decreases with the distance from the winner neuron. Therefore, the
adaption method is extended by a neighborhood function v (see also Fig. 3):

w~s 0 = w~s + v(i, s) · σ · (w~s − w

~i ),

where σ is a learning rate. By this learning procedure, the structure in the high-
dimensional sample data is non-linearly projected to the lower-dimensional topology.
After learning, arbitrary vectors (i.e. vectors from the sample set or prior ‘unknown’
vectors) can be propagated through the network and are mapped to the output units.

20
Figure 3: Network architecture of self-organizing maps (left) and possible neighbor-
hood function v for increasing distances from s (right)

For further details on self-organizing maps see [Koh84]. Examples for the application
of SOMs for text mining can be found in [LMS91, HKLK96, KKL+ 00, Nür01, RC01]
and in Sect. 3.4.2.

Model-based Clustering Using the EM-Algorithm Clustering can also be viewed

from a statistical point of view. If we have k different clusters we may either assign a
document di with certainty to a cluster (hard clustering) or assign di with probability qic
Pk
to Pc (soft clustering), where qi = (qi1 , . . . , qik ) is a probability vector c=1 qic = 1.
The underlying statistical assumption is that a document was created in two stages:
First we pick a cluster Pc from {1, . . . , k} with fixed probability qc ; then we generate
the words t of the document according to a cluster-specific probability distribution
p(t|Pc ). This corresponds to a mixture model where the probability of an observed
document (t1 , . . . , tni ) is
k
X
p(t1 , . . . , tni ) = qc p(t1 , . . . , tni |Pc ) (17)
c=1

Each cluster Pc is a mixture component. The mixture probabilities qc describe an un-

observable “cluster variable” z which may take the values from {1, . . . , k}. A well
established method for estimating models involving unobserved variables is the EM-
algorithm [HTF01], which basically replaces the unknown value with its current prob-
ability estimate and then proceeds as if it has been observed. Clustering methods for
documents based on mixture models have been proposed by Cheeseman [CS96] and
yield excellent results. Hofmann [Hof01] formulates a variant that is able to cluster
terms occurring together instead of documents.

3.2.3 Alternative Clustering Approaches

Co-clustering algorithm designate the simultaneous clustering of documents and
terms [DMM03]. They follow thereby another paradigm than the ”classical” cluster
algorithm as k-means which only clusters elements of the one dimension on the basis
of their similarity to the second one, e.g. documents based on terms.

21
Fuzzy Clustering While most classical clustering algorithms assign each datum to
exactly one cluster, thus forming a crisp partition of the given data, fuzzy clustering al-
lows for degrees of membership, to which a datum belongs to different clusters [Bez81].
These approaches are frequently more stable. Applications to text are described in, e.g.,
[MS01b, BN04].

The Utility of Clustering We have described the most important types of clustering
approaches, but we had to leave out many other. Obviously there are many ways to
define clusters and because of this we cannot expect to obtain something like the ‘true’
clustering. Still clustering can be insightful. In contrast to classification, which relies
on a prespecified grouping, cluster procedures label documents in a new way. By
studying the words and phrases that characterize a cluster, for example, a company
could learn new insights about its customers and their typical properties. A comparison
of some clustering methods is given in [SKK00].

3.3 Information Extraction

Natural language text contains much information that is not directly suitable for auto-
matic analysis by a computer. However, computers can be used to sift through large
amounts of text and extract useful information from single words, phrases or passages.
Therefore information extraction can be regarded as a restricted form of full natural
language understanding, where we know in advance what kind of semantic informa-
tion we are looking for. The main task is to extract parts of text and assign specific
attributes to it.
As an example consider the task to extract executive position changes from news
stories: ”Robert L. James, chairman and chief executive officer of McCann-Erickson,
is going to retire on July 1st. He will be replaced by John J. Donner, Jr., the agen-
cies chief operating officer.” In this case we have to identify the following information:
Organization (McCann-Erickson), position (chief executive officer), date (July 1), out-
going person name (Robert L. James), and incoming person name (John J. Donner,
Jr.).
The task of information extraction naturally decomposes into a series of processing
steps, typically including tokenization, sentence segmentation, part-of-speech assign-
ment, and the identification of named entities, i.e. person names, location names and
names of organizations. At a higher level phrases and sentences have to be parsed,
semantically interpreted and integrated. Finally the required pieces of information
like ”position” and ”incoming person name” are entered into the database. Although
the most accurate information extraction systems often involve handcrafted language-
processing modules, substantial progress has been made in applying data mining tech-
niques to a number of these steps.

3.3.1 Classification for Information Extraction

Entity extraction was originally formulated in the Message Understanding Conference
[Chi97]. One can regard it as a word-based tagging problem: The word, where the
entity starts, get tag ”B”, continuation words get tag ”I” and words outside the entity

22
get tag ”O”. This is done for each type of entity of interest. For the example above we
have for instance the person-words ”by (O) John (B) J. (I) Donner (I) Jr. (I) the (O)”.
Hence we have a sequential classification problem for the labels of each word, with
the surrounding words as input feature vector. A frequent way of forming the feature
vector is a binary encoding scheme. Each feature component can be considered as a test
that asserts whether a certain pattern occurs at a specific position or not. For example,
a feature component takes the value 1 if the previous word is the word ”John” and
0 otherwise. Of course we may not only test the presence of specific words but also
whether the words starts with a capital letter, has a specific suffix or is a specific part-
of-speech. In this way results of previous analysis may be used.
Now we may employ any efficient classification method to classify the word labels
using the input feature vector. A good candidate is the Support Vector Machine because
of its ability to handle large sparse feature vectors efficiently. [TC02] used it to extract
entities in the molecular biology domain.

3.3.2 Hidden Markov Models

One problem of standard classification approaches is that they do not take into account
the predicted labels of the surrounding words. This can be done using probabilistic
models of sequences of labels and features. Frequently used is the hidden Markov
model (HMM), which is based on the conditional distributions of current labels L(j)
given the previous label L(j−1) and the distribution of the current word t(j) given the
current and the previous labels L(j) , L(j−1) .
L(j) ∼ p(L(j) |L(j−1) ) t(j) ∼ p(t(j) |L(j) , L(j−1) ) (18)
A training set of words and their correct labels is required. For the observed words
the algorithm takes into account all possible sequences of labels and computes their
probabilities. An efficient learning method that exploits the sequential structure is the
Viterbi algorithm [Rab89]. Hidden Markov models were successfully used for named
entity extraction, e.g. in the Identifinder system [BSW99].

3.3.3 Conditional Random Fields

Hidden Markov models require the conditional independence of features of different
words given the labels. This is quite restrictive as we would like to include features
which correspond to several words simultaneously. A recent approach for modelling
this type of data is called conditional random field (CRF) [LMP01]. Again we consider
the observed vector of words t and the corresponding vector of labels L. The labels
have a graph structure. For a label Lc let N (c) be the indices of neighboring labels.
Then (t, L) is a conditional random field when conditioned on the vector t of all terms
the random variables obey the Markov property
p(Lc |t, Ld ; d 6= c) = p(Lc |t, Ld ; d ∈ N (c)) (19)
i.e. the whole vector t of observed terms and the labels of neighbors may influence
the distribution of the label Lc . Note that we do not model the distribution p(t) of the
observed words, which may exhibit arbitrary dependencies.

23
 We consider the simple case that the words t = (t1 , t2 , . . . , tn ) and the corre-
sponding labels L1 , L2 , . . . , Ln have a chain structure and that Lc depends only on
the preceding and succeeding labels Lc−1 and Lc+1 . Then the conditional distribution
p(L|t) has the form
 
n X kj n−1 mj
1 X XX
p(L|t) = exp  λjr fjr (Lj , t) + µjr gjr (Lj , Lj−1 , t) (20)
const
j=1 r=1 j=1 r=1

where fjr (Lj , t) and gjr (Lj , Lj−1 , t) are different features functions related to Lj
and the pair Lj , Lj−1 respectively. CRF models encompass hidden Markov models,
but they are much more expressive because they allow arbitrary dependencies in the
observation sequence and more complex neighborhood structures of labels. As for
most machine learning algorithms a training sample of words and the correct labels is
required. In addition to the identity of words arbitrary properties of the words, like
part-of-speech tags, capitalization, prefixes and suffixes, etc. may be used leading to
sometimes more than a million features. The unknown parameter values λjr and µjr
are usually estimated using conjugate gradient optimization routines [McC03].
McCallum [McC03] applies CRFs with feature selection to named entity recog-
nition and reports the following F1-measures for the CoNLL corpus: person names
93%, location names 92%, organization names 84%, miscellaneous names 80%. CRFs
also have been successfully applied to noun phrase identification [McC03], part-of-
speech tagging [LMP01], shallow parsing [SP03], and biological entity recognition
[KOT+ 04].

3.4 Explorative Text Mining: Visualization Methods

Graphical visualization of information frequently provides more comprehensive and
better and faster understandable information than it is possible by pure text based de-
scriptions and thus helps to mine large document collections. Many of the approaches
developed for text mining purposes are motivated by methods that had been proposed in
the areas of explorative data analysis, information visualization and visual data mining.
For an overview of these areas of research see, e.g., [UF01, Kei02]. In the following
we will focus on methods that have been specifically designed for text mining or — as
a subgroup of text mining methods and a typical application of visualization methods
— information retrieval.
In text mining or information retrieval systems visualization methods can improve
and simplify the discovery or extraction of relevant patterns or information. Informa-
tion that allow a visual representation comprises aspects of the document collection or
result sets, keyword relations, ontologies or — if retrieval systems are considered —
aspects of the search process itself, e.g. the search or navigation path in hyperlinked
collections.
However, especially for text collections we have the problem of finding an appro-
priate visualization for abstract textual information. Furthermore, an interactive visual
data exploration interface is usually desirable, e.g. to zoom in local areas or to select or
mark parts for further processing. This results in great demands on the user interface

24
and the hardware. In the following we give a brief overview of visualization methods
that have been realized for text mining and information retrieval systems.

3.4.1 Visualizing Relations and Result Sets

Interesting approaches to visualize keyword-document relations are, e.g., the Cat-a-
Cone model [HK97], which visualizes in a three dimensional representation hierarchies
of categories that can be interactively used to refine a search. The InfoCrystal [Spo95]
visualizes a (weighted) boolean query and the belonging result set in a crystal structure.
The lyberworld model [HKW94] and the visualization components of the SENTINEL
Model [FFKS99] are representing documents in an abstract keyword space.
An approach to visualize the results of a set of queries was presented in [HHP+ 01].
Here, retrieved documents are arranged according to their similarity to a query on
straight lines. These lines are arranged in a circle around a common center, i.e. ev-
ery query is represented by a single line. If several documents are placed on the same
(discrete) position, they are arranged in the same distance to the circle, but with a slight
offset. Thus, clusters occur that represent the distribution of documents for the belong-
ing query.

3.4.2 Visualizing Document Collections

For the visualization of document collections usually two-dimensional projections are
used, i.e. the high dimensional document space is mapped on a two-dimensional sur-
face. In order to depict individual documents or groups of documents usually text flags
are used, which represent either a keyword or the document category. Colors are fre-
quently used to visualize the density, e.g. the number of documents in this area, or the
difference to neighboring documents, e.g. in order to emphasize borders between dif-
ferent categories. If three-dimensional projections are used, for example, the number
of documents assigned to a specific area can be represented by the z-coordinate.

An Example: Visualization using Self-Organizing Maps Visualization of docu-

ment collections requires methods that are able to group documents based on their
similarity and furthermore that visualize the similarity between discovered groups of
documents. Clustering approaches that are frequently used to find groups of documents
with similar content [SKK00] – see also section 3.2 – usually do not consider the neigh-
borhood relations between the obtained cluster centers. Self-organizing maps, as dis-
cussed above, are an alternative approach which is frequently used in data analysis to
cluster high dimensional data. The resulting clusters are arranged in a low-dimensional
topology that preserves the neighborhood relations of the corresponding high dimen-
sional data vectors and thus not only objects that are assigned to one cluster are similar
to each other, but also objects of nearby clusters are expected to be more similar than
objects in more distant clusters.
Usually, two-dimensional arrangements of squares or hexagons are used for the
definition of the neighborhood relations. Although other topologies are possible for
self-organizing maps, two-dimensional maps have the advantage of intuitive visual-
ization and thus good exploration possibilities. In document retrieval, self-organizing

25
maps can be used to arrange documents based on their similarity. This approach opens
up several appealing navigation possibilities. Most important, the surrounding grid
cells of documents known to be interesting can be scanned for further similar docu-
ments. Furthermore, the distribution of keyword search results can be visualized by
coloring the grid cells of the map with respect to the number of hits. This allows a user
to judge e.g. whether the search results are assigned to a small number of (neighboring)
grid cells of the map, or whether the search hits are spread widely over the map and
thus the search was - most likely - too unspecific.
A first application of self-organizing maps in information retrieval was presented
in [LMS91]. It provided a simple two-dimensional cluster representation (categoriza-
tion) of a small document collection. A refined model, the WEBSOM approach, ex-
tended this idea to a web based interface applied to newsgroup data that provides simple
zooming techniques and coloring methods [HKLK96, Hon97, KKL+ 00]. Further ex-
tensions introduced hierarchies [Mer98], supported the visualization of search results
[RC01] and combined search, navigation and visualization techniques in an integrated
tool [Nür01]. A screenshot of the prototype discussed in [Nür01] is depicted in Fig. 4.

Figure 4: A Prototypical Retrieval System Based on Self-Organizing Maps

3.4.3 Other Techniques

Besides methods based on self-organizing maps several other techniques have been
successfully applied to visualize document collections. For example, the tool VxIn-
sight [BWD02] realizes a partially interactive mapping by an energy minimization ap-
proach similar to simulated annealing to construct a three dimensional landscape of the
document collection. As input either a vector space description of the documents or a
list of directional edges, e.g. defined based on citations of links, can be used. The tool

26
SPIRE [WTP+ 95] applies a three step approach: It first clusters documents in docu-
ment space, than projects the discovered cluster centers onto a two dimensional surface
and finally maps the documents relative to the projected cluster centers. SPIRE offers a
scatter plot like projection as well as a three dimensional visualization. The visualiza-
tion tool SCI-Map [Sma99] applies an iterative clustering approach to create a network
using, e.g., references of scientific publications. The tools visualizes the structure by a
map hierarchy with an increasing number of details.
One major problem of most existing visualization approaches is that they create
their output only by use of data inherent information, i.e. the distribution of the doc-
uments in document space. User specific information can not be integrated in order
to obtain, e.g., an improved separation of the documents with respect to user defined
criteria like keywords or phrases. Furthermore, the possibilities for a user to interact
with the system in order to navigate or search are usually very limited, e.g., to boolean
keyword searches and simple result lists.

3.5 Further Application Areas

Further major applications of text mining methods consider the detection of topics in
text streams and text summarization.
Topic detection studies the problem of detecting new and upcoming topics in time-
ordered document collections. The methods are frequently used in order to detect
and monitor (topic tracking) news tickers or news broadcasts. An introduction and
overview of current approaches can be found in [All02].
Text summarization aims at the creation of a condensed version of a document or a
document collection (multidocument summarization) that should contain its most im-
portant topics. Most approaches still focus on the idea to extract individual informative
sentences from a text. The summary consists then simply of a collection of these sen-
tences. However, recently refined approaches try to extract semantic information from
documents and create summaries based on this information (cf. [LGMF04]). For an
overview see [MM99] and [RHM02].

4 Applications
In this section we briefly discuss successful applications of text mining methods in
quite diverse areas as patent analysis, text classification in news agencies, bioinformat-
ics and spam filtering. Each of the applications has specific characteristics that had to
be considered while selecting appropriate text mining methods.

4.1 Patent Analysis

In recent years the analysis of patents developed to a large application area. The rea-
sons for this are on the one hand the increased number of patent applications and on the
other hand the progress that had been made in text classification, which allows to use
these techniques in this due to the commercial impact quite sensitive area. Meanwhile,
supervised and unsupervised techniques are applied to analyze patent documents and

27
to support companies and also the European patent office in their work. The challenges
in patent analysis consists of the length of the documents, which are larger then docu-
ments usually used in text classification, and the large number of available documents
in a corpus [KSB01]. Usually every document consist of 5000 words in average. More
than 140000 documents have to be handled by the European patent office (EPO) per
year. They are processed by 2500 patent examiners in three locations.
In several studies the classification quality of state-of-the-art methods was ana-
lyzed. [KSB01] reported very good result with an 3% error rate for 16000 full text
documents to be classified in 16 classes (mono-classification) and a 6% error rate in
the same setting for abstracts only by using the Winnow [Lit88] and the Rocchio algo-
rithm [Roc71]. These results are possible due to the large amount of available training
documents. Good results are also reported in [KZ02] for an internal EPO text classifi-
cation application with a precision of 81 % and an recall of 78 %.
Text clustering techniques for patent analysis are often applied to support the analy-
sis of patents in large companies by structuring and visualizing the investigated corpus.
Thus, these methods find their way in a lot of commercial products but are still also
of interest for research, since there is still a need for improved performance. Compa-
nies like IBM offer products to support the analysis of patent text documents. Dorre
describes in [DGS99] the IBM Intelligent Miner for text in a scenario applied to patent
text and compares it also to data mining and text mining. Coupet [CH98] does not only
apply clustering but also gives some nice visualization. A similar scenario on the basis
of SOM is given in [LAHF03].

4.2 Text Classification for News Agencies

In publishing houses a large number of news stories arrive each day. The users like
to have these stories tagged with categories and the names of important persons, or-
ganizations and places. To automate this process the Deutsche Presse-Agentur (dpa)
and a group of leading German broadcasters (PAN) wanted to select a commercial text
classification system to support the annotation of news articles. Seven systems were
tested with a two given test corpora of about half a million news stories and different
categorical hierarchies of about 800 and 2300 categories [Pd05]. Due to confidentiality
the results can be published only in anonymized form.
For the corpus with 2300 categories the best system achieved at an F1-value of
39%, while for the corpus with 800 categories an F1-value of 79% was reached. In the
latter case a partially automatic assignment based on the reliability score was possible
for about half the documents, while otherwise the systems could only deliver proposals
for human categorizers. Especially good are the results for recovering persons and
geographic locations with about 80% F1-value. In general there were great variations
between the performances of the systems.
In a usability experiment with human annotators the formal evaluation results were
confirmed leading to faster and more consistent annotation. It turned out, that with
respect to categories the human annotators exhibit a relative large disagreement and a
lower consistency than text mining systems. Hence the support of human annotators by
text mining systems offers more consistent annotations in addition to faster annotation.

28
The Deutsche Presse-Agentur now is routinely using a text mining system in its news
production workflow.

4.3 Bioinformatics
Bio-entity recognition aims to identify and classify technical terms in the domain of
molecular biology that correspond to instances of concepts that are of interest to biolo-
gists. Examples of such entities include the names of proteins, genes and their locations
of activity such as cells or organism names. Entity recognition is becoming increas-
ingly important with the massive increase in reported results due to high throughput
experimental methods. It can be used in several higher level information access tasks
such as relation extraction, summarization and question answering.
Recently the GENIA corpus was provided as a benchmark data set to compare
different entity extraction approaches [KOT+ 04]. It contains 2000 abstracts from the
MEDLINE database which were hand annotated with 36 types of biological entities.
The following sentence is an example: “We have shown that <protein> interleukin-
1 </protein> (<protein> IL-1 </protein>) and <protein> IL-2 </protein> control
<DNA> IL-2 receptor alpha (IL-2R alpha) gene </DNA> transcription in <cell line>
CD4-CD8- murine T lymphocyte precursors </cell line>”.
In the 2004 evaluation four types of extraction models were used: Support Vec-
tor Machines (SVMs), Hidden Markov Models (HMMs), Conditional Random Fields
(CRFs) and the related Maximum Entropy Markov Models (MEMMs). Varying types
of input features were employed: lexical features (words), n-grams, orthographic in-
formation, word lists, part-of-speech tags, noun phrase tags, etc. The evaluation shows
that the best five systems yield an F1-value of about 70% [KOT+ 04]. They use SVMs
in combination with Markov models (72.6%), MEMMs (70.1%), CRFs (69.8%), CRFs
together with SVMs (66.3%), and HMMs (64.8%). For practical applications the cur-
rent accuracy levels are not yet satisfactory and research currently aims at including
a sophisticated mix of external resources such as keyword lists and ontologies which
provide terminological resources.

4.4 Anti-Spam Filtering of Emails

The explosive growth of unsolicited e-mail, more commonly known as spam, over the
last years has been undermining constantly the usability of e-mail. One solution is
offered by anti-spam filters. Most commercially available filters use black-lists and
hand-crafted rules. On the other hand, the success of machine learning methods in
text classification offers the possibility to arrive at anti-spam filters that quickly may be
adapted to new types of spam.
There is a growing number of learning spam filters mostly using naive Bayes clas-
sifiers. A prominent example is Mozilla’s e-mail client. Michelakis et al. [MAP+ 04]
compare different classifier methods and investigate different costs of classifying a
proper mail as spam. They find that for their benchmark corpora the SVM nearly al-
ways yields best results.
To explore how well a learning-based filter performs in real life, they used an SVM-
based procedure for seven months without retraining. They achieved a precision of

29
96.5% and a recall of 89.3%. They conclude that these good results may be improved
by careful preprocessing and the extension of filtering to different languages.

5 Conclusion
In this article, we tried to give a brief introduction to the broad field of text mining.
Therefore, we motivated this field of research, gave a more formal definition of the
terms used herein and presented a brief overview of currently available text mining
methods, their properties and their application to specific problems. Even though, it
was impossible to describe all algorithms and applications in detail within the (size)
limits of an article, we think that the ideas discussed and the provided references should
give the interested reader a rough overview of this field and several starting points for
further studies.

References
[Abn91] S. P. Abney. Parsing by chunks. In R. C. Berwick, S. P. Abney, and
C. Tenny, editors, Principle-Based Parsing: Computation and Psycholin-
guistics, pages 257–278. Kluwer Academic Publishers, Boston, 1991.

[All02] J. Allan, editor. Topic Detection and Tracking. Kluwer Academic Pub-
lishers, Norwell, MA, 2002.

[Be99] M. B. and D. J. Hand (eds.). Intelligent data analysis. Springer-Verlag

New York, Inc., 1999.

[Bez81] J. C. Bezdek. Pattern Recognition with Fuzzy Objective Function Algo-

rithms. Plenum Press, New York, 1981.

[BH04] S. Bloehdorn and A. Hotho. Text classification by boosting weak learners

based on terms and concepts. In Proc. IEEE Int. Conf. on Data Mining
(ICDM 04), pages 331–334. IEEE Computer Society Press, NOV 2004.
[BN04] C. Borgelt and A. Nürnberger. Fast fuzzy clustering of web page col-
lections. In Proc. of PKDD Workshop on Statistical Approaches for Web
Mining (SAWM), Pisa, Italy, 2004.
[BSW99] D. Bikel, R. Schwartz, and R. Weischedel. An algorithm that learns what’s
in a name. Machine learning, 34:211–231, 1999.
[BWD02] K. W. Boyack, B. N. Wylie, and G. S. Davidson. Domain visualization
using vxinsight for science and technology management. Journal of the
American Society for Information Science and Technologie, 53(9):764–
774, 2002.
[BYRN99] R. Baeza-Yates and B. Ribeiro-Neto. Modern Information Retrieval. Ad-
dison Wesley Longman, 1999.

30
[CH98] P. Coupet and M. Hehenberger. Text mining applied to patent analysis.
In Annual Meeting of American Intellectual Property Law Association
(AIPLA) Airlington, 1998.
[Chi97] N. Chinchor. Muc-7 named entity task definition version 3.5. Technical
report, NIST, ftp.muc.saic.com/pub/MUC/MUC7-guidelines, 1997.
[CHY96] M.-S. Chen, J. Han, and P. S. Yu. Data mining: an overview from a
database perspective. IEEE Transaction on Knowledge and Data Engi-
neering, 8(6):866–883, 1996.
[cri99] Cross industry standard process for data mining. http://www.
crisp-dm.org/, 1999.
[CS96] P. Cheeseman and J. Stutz. Bayesian classification (AutoClass): Theory
and results. In U.M. Fayyad, G. Piatetsky-Shapiro, P. Smyth, and R. Uthu-
rusamy, editors, Advances in Knowledge Discovery and Data Mining,
pages 153–180. AAAI/MIT Press, 1996.
[DDFL90] S. Deerwester, S.T. Dumais, G.W. Furnas, and T.K. Landauer. Indexing
by latent semantic analysis. Journal of the American Society for Informa-
tion Sciences, 41:391–407, 1990.
[DGS99] J. Dörre, P. Gerstl, and R. Seiffert. Text mining: finding nuggets in moun-
tains of textual data. In Proc. 5th ACM Int. Conf. on Knowledge Discovery
and Data Mining (KDD-99), pages 398–401, San Diego, US, 1999. ACM
Press, New York, US.

[DH73] R. O. Duda and P. E. Hart. Pattern Classification and Scene Analysis.

J. Wiley & Sons, New York, NY, USA, 1973.

[DLR77] A. P. Dempster, N. M. Laird, and D. B. Rubin. Maximum likelihood

from incomplete data via the em algorithm. Journal of the Royal Statistic
Society, Series B, 39(1):1–38, 1977.

[DMM03] I.S. Dhillon, S. Mallela, and D.S. Modha. Information-theoretic co-

clustering. In Proc. of the ninth ACM SIGKDD int. conf. on Knowledge
Discovery and Data Mining, pages 89–98. ACM Press, 2003.

[DPHS98] S. Dumais, J. Platt, D. Heckerman, and M. Sahami. Inductive learning

algorithms and representations for text categorization. In 7th Int. Conf.
on Information and Knowledge Managment, 1998.
[FBY92] W. B. Frakes and R. Baeza-Yates. Information Retrieval: Data Structures
& Algorithms. Prentice Hall, New Jersey, 1992.
[FD95] R. Feldman and I. Dagan. Kdt - knowledge discovery in texts. In Proc.
of the First Int. Conf. on Knowledge Discovery (KDD), pages 112–117,
1995.

31
[FFKS99] K. L. Fox, O. Frieder, M. M. Knepper, and E. J. Snowberg. Sentinel: A
multiple engine information retrieval and visualization system. Journal
of the American Society of Information Science, 50(7):616–625, 1999.
[Fic97] N. Fickel. Clusteranalyse mit gemischt-skalierten merkmalen: Ab-
strahierung vom skalenniveau. Allg. Statistisches Archiv, 81(3):249–265,
1997.
[For65] E. Forgy. Cluster analysis of multivariate data: Efficiency versus inter-
pretability of classification. Biometrics, 21(3):768–769, 1965.
[FPSS96] U. M. Fayyad, G. Piatetsky-Shapiro, and P. Smyth. Knowledge discovery
and data mining: Towards a unifying framework. In Knowledge Discov-
ery and Data Mining, pages 82–88, 1996.
[Gai03] R. Gaizauskas. An information extraction perspective on
text mining: Tasks, technologies and prototype applications.
http://www.itri.bton.ac.uk/projects/euromap/
TextMiningEvent/Rob_Gaizauskas.pdf, 2003.
[GG92] A. Gersho and R. M. Gray. Vector quantization and signal compression.
Kluwer Academic Publishers, 1992.
[Goo65] I. J. Good. The Estimation of Probabilities: An Essay on Modern
Bayesian Methods. MIT Press, Cambridge, MA, 1965.
[Gre98] W. R. Greiff. A theory of term weighting based on exploratory data anal-
ysis. In 21st Annual International ACM SIGIR Conference on Research
and Development in Information Retrieval, New York, NY, 1998. ACM.
[Har75] J. Hartigan. Clustering Algorithms. John Wiley and Sons, New York,
1975.
[Hea99] M. Hearst. Untangling text data mining. In Proc. of ACL’99 the 37th
Annual Meeting of the Association for Computational Linguistics, 1999.
[HHP+ 01] S. Havre, E. Hetzler, K. Perrine, E. Jurrus, and N. Miller. Interactive
visualization of multiple query result. In Proc. of IEEE Symposium on
Information Visualization 2001, pages 105 –112. IEEE, 2001.
[Hid02] J. M. G. Hidalgo. Tutorial on text mining and internet content filter-
ing. Tutorial Notes Online: http://ecmlpkdd.cs.helsinki.
fi/pdf/hidalgo.pdf, 2002.
[HK97] M. A. Hearst and C. Karadi. Cat-a-cone: An interactive interface for
specifying searches and viewing retrieval results using a large category
hierarchie. In Proc. of the 20th Annual Int. ACM SIGIR Conference, pages
246–255. ACM, 1997.

32
[HKLK96] T. Honkela, S. Kaski, K. Lagus, and T. Kohonen. Newsgroup explo-
ration with the websom method and browsing interface, technical report.
Technical report, Helsinki University of Technology, Neural Networks
Research Center, Espoo, Finland, 1996.
[HKW94] M. Hemmje, C. Kunkel, and A. Willett. Lyberworld - a visualization user
interface supporting fulltext retrieval. In Proc. of ACM SIGIR 94, pages
254–259. ACM, 1994.
[Hof01] T. Hofmann. Unsupervised learning by probabilistic latent semantic anal-
ysis. Machine Learning Journal, 41(1):177–196, 2001.
[Hon97] T. Honkela. Self-Organizing Maps in Natural Language Processing. PhD
thesis, Helsinki Univ. of Technology, Neural Networks Research Center,
Espoo, Finland, 1997.
[HSS03] A. Hotho, S. Staab, and G. Stumme. Ontologies improve text document
clustering. In Proc. IEEE Int. Conf. on Data Mining (ICDM 03), pages
541–544, 2003.
[HTF01] T. Hastie, R. Tibshirani, and J. Friedman. The Elements of Statistical
Learning. Springer, 2001.
[Joa98] T. Joachims. Text categorization with support vector machines: Learning
with many relevant features. In C. Nedellec and C. Rouveirol, editors,
European Conf. on Machine Learning (ECML), 1998.
[Kei02] D. A. Keim. Information visualization and visual data mining. IEEE
Transactions on Visualization and Computer Graphics, 7(2):100–107,
2002.
[KJ03] V. Kumar and M. Joshi. What is data mining? http://www-users.
cs.umn.edu/˜mjoshi/hpdmtut/sld004.htm, 2003.
[KKL+ 00] T. Kohonen, S. Kaski, K. Lagus, J. Salojärvi, J. Honkela, V. Paattero, and
A. Saarela. Self organization of a massive document collection. IEEE
Transactions on Neural Networks, 11(3):574–585, May 2000.
[Kod99] Y. Kodratoff. Knowledge discovery in texts: A definition and applica-
tions. Lecture Notes in Computer Science, 1609:16–29, 1999.
[Koh82] T. Kohonen. Self-organized formation of topologically correct feature
maps. Biological Cybernetics, 43:59–69, 1982.
[Koh84] T. Kohonen. Self-Organization and Associative Memory. Springer-
Verlag, Berlin, 1984.
[KOT+ 04] J. Kim, T. Ohta, Y. Tsuruoka, Y. Tateisi, and N. Collier. Introduction to
the bio-entity task at jnlpba. In N. Collier, P. Ruch, and A. Nazarenko,
editors, Proc. Workshop on Natural Language Processing in Biomedicine
and its Applications, pages 70–76, 2004.

33
[KR90] L. Kaufman and P. J. Rousseeuw. Finding groups in data: an introduction
to cluster analysis. Wiley, New York, 1990.

[KSB01] C.H.A. Koster, M. Seutter, and J. Beney. Classifying patent applications

with winnow. In Proceedings Benelearn, Antwerpen, 2001.

[KZ02] M. Krier and F. Zacca. Automatic categorisation applications at the euro-

pean patent office. World Patent Information, 24(3):187–196, September
2002.
[LAHF03] J.-C. Lamirel, S. Al Shehabi, M. Hoffmann, and C. Francois. Intelligent
patent analysis through the use of a neural network: Experiment of multi-
viewpoint analysis with the multisom model. In ACL-2003 Workshop on
Patent Corpus Processing, 2003.
[LGMF04] Jure Leskovec, Marko Grobelnik, and Natasa Milic-Frayling. Learning
sub-structures of document semantic graphs for document summariza-
tion. In KDD 2004 Workshop on Link Analysis and Group Detection
(LinkKDD), Seattle, Washington, 2004.
[Lit88] N. Littlestone. Learning quickly when irrelevant attributes abound: A
new linear-threshold algorithm. Machine Learning, 2(4):285–318, 1988.
[LK02] E. Leopold and J. Kindermann. Text categorization with support vector
machines. How to represent texts in input space? Machine Learning,
46:423 – 444, 2002.
[LMP01] J. Lafferty, A. McCallum, and F. Pereira. Conditional random fields:
Probabilistic models for segmenting and labeling sequence data. In Proc.
ICML, 2001.

[LMS91] X. Lin, G. Marchionini, and D. Soergel. A selforganizing seman-

tic map for information retrieval. In Proc. of the 14th International
ACM/SIGIR Conference on Research and Development in Information
Retrieval, pages 262–269, New York, 1991. ACM Press.

[LS89] K. E. Lochbaum and L. A. Streeter. Combining and comparing the ef-

fectiveness of latent semantic indexing and the ordinary vector space
model for information retrieval. Information Processing and Manage-
ment, 25(6):665–676, 1989.
[Mac67] J. B. MacQueen. Some methods for classification and analysis of multi-
variate observations. In L. M. Le Cam and J. Neyman, editors, Proc. of
the fifth Berkeley Symposium on Mathematical Statistics and Probability,
volume 1, pages 281–297. University of California Press, 1967.
[Mai02] R. Maitra. A statistical perspective on data mining. J. Ind. Soc. Prob.
Statist., 2002.

34
[MAP+ 04] E. Michelakis, I. Androutsopoulos, G. Paliouras, G. Sakkis, and P. Stam-
atopoulos. Filtron: A learning-based anti-spam filter. In Proc. 1st Conf.
on Email and Anti-Spam (CEAS 2004), Mountain View, CA, USA, 2004.
[McC03] A. McCallum. Efficiently inducing features of conditional random fields.
In Proc. Conf. on Uncertainty in Articifical Intelligence (UAI), 2003.,
2003.
[Mer98] D. Merkl. Text classification with self-organizing maps: Some lessons
learned. Neurocomputing, 21:61–77, 1998.
[Mit97] T. Mitchell. Machine Learning. McGraw-Hill, 1997.
[MM99] I. Mani and M. T. Maybury, editors. Advances in Automatic Text Summa-
rization. MIT Press, 1999.
[MS01a] C. D. Manning and H. Schütze. Foundations of Statistical Natural Lan-
guage Processing. MIT Press, Cambridge, MA, 2001.
[MS01b] M. E. Mendes and Lionel Sacks. Dynamic knowledge representation
for e-learning applications. In Proc. of BISC International Workshop on
Fuzzy Logic and the Internet (FLINT 2001), pages 176–181, Berkeley,
USA, 2001. ERL, College of Engineering, University of California.
[NM02] U. Nahm and R. Mooney. Text mining with information extraction. In
Proceedings of the AAAI 2002 Spring Symposium on Mining Answers
from Texts and Knowledge Bases, 2002.

[NMTM00] K. Nigam, A. McCallum, S. Thrun, and T. Mitchell. Text classification

from labeled and unlabeled documents using em. Machine Learning,
39:103–134, 2000.
[Nür01] A. Nürnberger. Interactive text retrieval supported by growing self-
organizing maps. In Timo Ojala, editor, Proc. of the International Work-
shop on Information Retrieval (IR 2001), pages 61–70, Oulu, Finland,
Sep 2001. Infotech.
[Pd05] G. Paaß and H. deVries. Evaluating the performance of text mining sys-
tems on real-world press archives. In Proc. 29th Annual Conference of
the German Classification Society (GfKl 2005). Springer, 2005.
[PL02] P. Pantel and D. Lin. Document clustering with committees. In Proc. of
SIGIR’02, Tampere, Finland, 2002.
[Por80] M. Porter. An algorithm for suffix stripping. Program, pages 130–137,
1980.
[Qui86] J. R. Quinlan. Induction of decision trees. Machine Learning, 1:81–106,
1986.

35
[Rab89] L. R. Rabiner. A tutorial on hidden markov models and selected applica-
tions in speech recognition. Proc. of IEEE, 77(2):257–286, 1989.

[RC01] D. G. Roussinov and H. Chen. Information navigation on the web by clus-

tering and summarizing query results. Information Processing & Man-
agement, 37(6):789–816, 2001.
[RHM02] D.R. Radev, E. Hovy, and K. McKeown. Introduction to the special issue
on summarization. Computational Linguistics, 28(4):399–408, 2002.
[Rob77] S. E. Robertson. The probability ranking principle. Journal of Documen-
tation, 33:294–304, 1977.
[Roc71] J. J. Rochio. Relevance feedback in information retrieval. In G. Salton,
editor, The SMART Retrieval System, pages 313–323. Prentice Hall, En-
glewood Cliffs, NJ, 1971.
[SAB94] G. Salton, J. Allan, and C. Buckley. Automatic structuring and retrieval
of large text files. Communications of the ACM, 37(2):97–108, Feb 1994.

[SB88] G. Salton and C. Buckley. Term weighting approaches in automatic text

retrieval. Information Processing & Management, 24(5):513–523, 1988.

[Seb02] F. Sebastiani. Machine learning in automated text categorization. ACM

Computing Surveys, 34:1–47, 2002.
[SEK03] M. Steinbach, L. Ertoz, and V. Kumar. Challenges of clustering high
dimensional data. In L. T. Wille, editor, New Vistas in Statistical Physics
– Applications in Econophysics, Bioinformatics, and Pattern Recognition.
Springer-Verlag, 2003.

[SJW97] K. Sparck-Jones and P. Willett, editors. Readings in Information Re-

trieval. Morgan Kaufmann, 1997.

[SKK00] M. Steinbach, Ge. Karypis, and V. Kumara. A comparison of document

clustering techniques. In KDD Workshop on Text Mining, 2000. (see also
TR 00-034, University of Minnesota, MN).
[Sma99] H. Small. Visualizing science by citation mapping. Journal of the Amer-
ican Society for Information Science, 50(9):799–813, 1999.
[SP03] F. Sha and F. Pereira. Shallow parsing with conditional random fields. In
Proc. Human Language Technology NAACL, 2003.
[Spo95] A. Spoerri. InfoCrystal: A Visual Tool for Information Retrieval. PhD
thesis, Massachusetts Institute of Technology, Cambridge, MA, 1995.
[SS99] R. E. Schapire and Y. Singer. Improved boosting using confidence-rated
predictions. Machine Learning, 37(3):297–336, 1999.

36
[SS00] R. E. Schapire and Y. Singer. BoosTexter: A boosting-based system for
text categorization. Machine Learning, 39(2/3):135–168, 2000.

[SWY75] G. Salton, A. Wong, and C. S. Yang. A vector space model for automatic
indexing. Communications of the ACM, 18(11):613–620, 1975. (see also
TR74-218, Cornell University, NY, USA).
[TC02] K. Takeuchi and N. Collier. Use of support vector machines in extended
named entity recognition. In 6th Conf. on Natural Language Learning
(CoNLL-02), pages 119–125, 2002.
[TMS05] Text mining summit conference brochure. http://www.
textminingnews.com/, 2005.
[UF01] A. Wierse U. Fayyad, G. Grinstein. Information Visualization in Data
Mining and Knowledge Discovery. Morgan Kaufmann, 2001.
[van86] C. J. van Rijsbergen. A non-classical logic for information retrieval. The
Computer Journal, 29(6):481–485, 1986.

[Wil97] Y. Wilks. Information extraction as a core language technology. In M-T.

Pazienza, editor, Information Extraction. Springer, Berlin, 1997.

[WMB99] I. H. Witten, A. Moffat, and T. C. Bell. Managing Gigabytes: Compress-

ing and Indexing Documents and Images. Morgan Kaufmann Publishers,
San Francisco, 1999.
[WTP+ 95] J. A. Wise, J. J. Thomas, K. Pennock, D. Lantrip, M. Pottier, A. Schur,
and V. Crow. Visualizing the non-visual: Spatial analysis and interac-
tion with information from text documents. In Proc. of IEEE Symposium
on Information Visualization ’95, pages 51–58. IEEE Computer Society
Press, 1995.

37