No Aprendizado de Máquina Supervisionado - AM - é apresentado ao algoritmo de indução um conjunto de instâncias de treinamento, no qual cada instância é um vetor de features rotulado com a classe. O algoritmo de indução tem como tarefa induzir um classificador que será utilizado para classificar novas instâncias. Algoritmos de indução convencionais baseam-se nos dados fornecidos pelo usuário para construir as descrições dos conceitos. Uma representação inadequada do espaço de busca ou da linguagem de descrição do conjunto de instâncias, bem como erros nos exemplos de treinamento, podem tornar os problemas de aprendizado difícies. Um dos problemas centrais em AM é a Seleção de um Subconjunto de Features - SSF - na qual o objetivo é tentar diminuir o número de features que serão fornecidas ao algoritmo de indução. São várias as razões para a realização de SSF. A primeira é que a maioria dos algoritmos de AM, computacionalmente viáveis, não trabalham bem na presença de muitas features, isto é a precisão dos classificadores gerados pode ser melhorada com a aplicação de SSF. Ainda, com um número menor de features, a compreensibilidade do conceito induzido pode ser melhorada. Uma terceira razão é o alto custo para coletar e processar grande quantidade de dados. Existem, basicamente, três abordagens para a SSF: embedded, filtro e wrapper. Por outro lado, se as features utilizadas para descrever os exemplos de treinamento são inadequadas, os algoritmos de aprendizado estão propensos a criar descrições excessivamente complexas e imprecisas. Porém, essas features, individualmente inadequadas, podem algumas vezes serem, convenientemente, combinadas gerando novas features que podem mostrar-se altamente representativas para a descrição de um conceito. O processo de construção de novas features é conhecido como Construção de Features ou Indução Construtiva - IC. Neste trabalho são enfocadas as abordagens filtro e wrapper para a realização de SSF, bem como a IC guiada pelo conhecimento. É descrita uma série de experimentos usando SSF e IC utilizando quatro conjuntos de dados naturais e diversos algoritmos simbólicos de indução. Para cada conjunto de dados e cada indutor, são realizadas várias medidas, tais como, precisão, tempo de execução do indutor e número de features selecionadas pelo indutor. São descritos também diversos experimentos realizados utilizando três conjuntos de dados do mundo real. O foco desses experimentos não está somente na avaliação da performance dos algoritmos de indução, mas também na avaliação do conhecimento extraído. Durante a extração de conhecimento, os resultados foram apresentados aos especialistas para que fossem feitas sugestões para experimentos futuros. Uma parte do conhecimento extraído desses três estudos de casos foram considerados muito interessantes pelos especialistas. Isso mostra que a interação de diferentes áreas de conhecimento, neste caso específico, áreas médica e computacional, pode produzir resultados interessantes. Assim, para que a aplicação do Aprendizado de Máquina possa gerar frutos é necessário que dois grupos de pesquisadores sejam unidos: aqueles que conhecem os métodos de AM existentes e aqueles com o conhecimento no domínio da aplicação para o fornecimento de dados e a avaliação do conhecimento adquirido. / In supervised Machine Learning - ML - an induction algorithm is typically presented with a set of training instances, where each instance is described by a vector of feature values and a class label. The task of the induction algorithm (inducer) is to induce a classifier that will be useful in classifying new cases. Conventional inductive-learning algorithms rely on existing (user) provided data to build their descriptions. Inadequate representation space or description language as well as errors in training examples can make learning problems be difficult. One of the main problems in ML is the Feature Subset Selection - FSS - problem, i.e. the learning algorithm is faced with the problem of selecting some subset of features upon which to focus its attention, while ignoring the rest. There are a variety of reasons that justify doing FSS. The first reason that can be pointed out is that most of the ML algorithms, that are computationally feasible, do not work well in the presence of a very large number of features. This means that FSS can improve the accuracy of the classifiers generated by these algorithms. Another reason to use FSS is that it can improve comprehensibility, i.e. the human ability of understanding the data and the rules generated by symbolic ML algorithms. A third reason for doing FSS is the high cost in some domains for collecting data. Finally, FSS can reduce the cost of processing huge quantities of data. Basically, there are three approaches in Machine Learning for FSS: embedded, filter and wrapper approaches. On the other hand, if the provided features for describing the training examples are inadequate, the learning algorithms are likely to create excessively complex and inaccurate descriptions. These individually inadequate features can sometimes be combined conveniently, generating new features which can turn out to be highly representative to the description of the concept. The process of constructing new features is called Constructive Induction - CI. Is this work we focus on the filter and wrapper approaches for FSS as well as Knowledge-driven CI. We describe a series of experiments for FSS and CI, performed on four natural datasets using several symbolic ML algorithms. For each dataset, various measures are taken to compare the inducers performance, for example accuracy, time taken to run the inducers and number of selected features by each evaluated induction algorithm. Several experiments using three real world datasets are also described. The focus of these three case studies is not only comparing the induction algorithms performance, but also the evaluation of the extracted knowledge. During the knowledge extraction step results were presented to the specialist, who gave many suggestions for the development of further experiments. Some of the knowledge extracted from these three real world datasets were found very interesting by the specialist. This shows that the interaction between different areas, in this case, medical and computational areas, may produce interesting results. Thus, two groups of researchers need to be put together if the application of ML is to bear fruit: those that are acquainted with the existing ML methods, and those with expertise in the given application domain to provide training data.

Redução de dimensionalidade é um problema muito importante da área de reconhecimento de padrões com aplicação em diversos campos do conhecimento. Dentre as técnicas de redução de dimensionalidade, a de seleção de características foi o principal foco desta pesquisa. De uma forma geral, a maioria dos métodos de redução de dimensionalidade presentes na literatura costumam privilegiar casos nos quais os dados sejam linearmente separáveis e só existam duas classes distintas. No intuito de tratar casos mais genéricos, este trabalho propõe uma função critério, baseada em sólidos princípios de teoria estatística como entropia e informação mútua, a ser embutida nos algoritmos de seleção de características existentes. A proposta dessa abordagem é tornar possível classificar os dados, linearmente separáveis ou não, em duas ou mais classes levando em conta um pequeno subespaço de características. Alguns resultados com dados sintéticos e dados reais foram obtidos confirmando a utilidade dessa técnica. Este trabalho tratou dois problemas de bioinformática. O primeiro trata de distinguir dois fenômenos biológicos através de seleção de um subconjunto apropriado de genes. Foi estudada uma técnica de seleção de genes fortes utilizando máquinas de suporte vetorial (MSV) que já vinha sendo aplicada para este fim em dados de SAGE do genoma humano. Grande parte dos genes fortes encontrados por esta técnica para distinguir tumores de cérebro (glioblastoma e astrocytoma), foram validados pela metodologia apresentada neste trabalho. O segundo problema que foi tratado neste trabalho é o de identificação de redes de regulação gênica, utilizando a metodologia proposta, em dados produzidos pelo trabalho de DeRisi et al sobre microarray do genoma do Plasmodium falciparum, agente causador da malária, durante as 48 horas de seu ciclo de vida. O presente texto apresenta evidências de que a utilização da entropia condicional média para estimar redes genéticas probabilísticas (PGN) pode ser uma abordagem bastante promissora nesse tipo de aplicação. No contexto de processamento de imagens, tal técnica pôde ser aplicada com sucesso em obter W-operadores minimais para realização de filtragem de imagens e reconhecimento de texturas. / Dimensionality reduction is a very important pattern recognition problem with many applications. Among the dimensionality reduction techniques, feature selection was the main focus of this research. In general, most dimensionality reduction methods that may be found in the literature privilegiate cases in which the data is linearly separable and with only two distinct classes. Aiming at covering more generic cases, this work proposes a criterion function, based on the statistical theory principles of entropy and mutual information, to be embedded in the existing feature selection algorithms. This approach allows to classify the data, linearly separable or not, in two or more classes, taking into account a small feature subspace. Results with synthetic and real data were obtained corroborating the utility of this technique. This work addressed two bioinformatics problems. The first is about distinguishing two biological fenomena through the selection of an appropriate subset of genes. We studied a strong genes selection technique using support vector machines (SVM) which has been applied to SAGE data of human genome. Most of the strong genes found by this technique to distinguish brain tumors (glioblastoma and astrocytoma) were validated by the proposed methodology presented in this work. The second problem covered in this work is the identification of genetic network regulation, using our proposed methodology, from data produced by work of DeRisi et al about microarray of the Plasmodium falciparum genome, malaria agent, during 48 hours of its life cycle. This text presents evidences that using mean conditional entropy to estimate a probabilistic genetic network (PGN) may be very promising. In the image processing context, it is shown that this technique can be applied to obtain minimal W-operators that perform image filtering and texture recognition.

This doctoral thesis is the result of a research effort performed in two fields of speech technology, i.e., speech recognition and mispronunciation detection. Although the two areas are clearly distinguishable, the proposed approaches share a common hypothesis based on psychoacoustic processing of speech signals. The conjecture implies that the human auditory periphery provides a relatively good separation of different sound classes. Hence, it is possible to use recent findings from psychoacoustic perception together with mathematical and computational tools to model the auditory sensitivities to small speech signal changes. The performance of an automatic speech recognition system strongly depends on the representation used for the front-end. If the extracted features do not include all relevant information, the performance of the classification stage is inherently suboptimal. The work described in Papers A, B and C is motivated by the fact that humans perform better at speech recognition than machines, particularly for noisy environments. The goal is to make use of knowledge of human perception in the selection and optimization of speech features for speech recognition. These papers show that maximizing the similarity of the Euclidean geometry of the features to the geometry of the perceptual domain is a powerful tool to select or optimize features. Experiments with a practical speech recognizer confirm the validity of the principle. It is also shown an approach to improve mel frequency cepstrum coefficients (MFCCs) through offline optimization. The method has three advantages: i) it is computationally inexpensive, ii) it does not use the auditory model directly, thus avoiding its computational cost, and iii) importantly, it provides better recognition performance than traditional MFCCs for both clean and noisy conditions. The second task concerns automatic pronunciation error detection. The research, described in Papers D, E and F, is motivated by the observation that almost all native speakers perceive, relatively easily, the acoustic characteristics of their own language when it is produced by speakers of the language. Small variations within a phoneme category, sometimes different for various phonemes, do not change significantly the perception of the language’s own sounds. Several methods are introduced based on similarity measures of the Euclidean space spanned by the acoustic representations of the speech signal and the Euclidean space spanned by an auditory model output, to identify the problematic phonemes for a given speaker. The methods are tested for groups of speakers from different languages and evaluated according to a theoretical linguistic study showing that they can capture many of the problematic phonemes that speakers from each language mispronounce. Finally, a listening test on the same dataset verifies the validity of these methods. / <p>QC 20120914</p> / European Union FP6-034362 research project ACORNS / Computer-Animated language Teachers (CALATea)

The availability of new technologies supplied life scientists with large amounts of experimental data. The data sets are large not only in terms of the number of observations, but also in terms of the number of recorded features. One of the aims of modeling is to explain a given phenomenon in possibly the simplest way, hence the need for selection of suitable features. We extended a Monte Carlo-based approach to selecting statistically significant features with discovery of feature interdependencies and used it in modeling sequence-function relationships in proteins. Our approach led to compact and easy-to-interpret predictive models. First, we represented protein sequences in terms of their physicochemical properties. This was followed by our feature selection and discovery of feature interdependencies. Finally, predictive models based on e.g., decision trees or rough sets were constructed. We applied the method to model two important biological problems: 1) HIV-1 resistance to reverse transcriptase-targeted drugs and 2) post-translational modifications of proteins. In the case of HIV resistance, we were not only able to predict whether the mutated protein is resistant to a drug or not, but we also suggested some new, previously neglected, mutations that possibly contribute to drug resistance. For all these mutations we proposed probable molecular mechanisms of action using literature and 3D structure studies. In the case of predicting PTMs, we built high accuracy models of modifications. In comparison to other methods, we were able to resolve whether the closest neighborhood of a residue (the nanomer) is sufficient to determine its modification status. Importantly, the application of our method yields networks of interdependent physicochemical properties of amino acids that show how these properties collaborate in establishing a given modification. We believe that the presented methods will help researchers to analyze a large class of important biological problems and will guide them in their research.

In this thesis, inference of biological networks from in vivo data generated by perturbation experiments is considered, i.e. deduction of causal interactions that exist among the observed variables. Knowledge of such regulatory influences is essential in biology. A system property–interampatteness–is introduced that explains why the variation in existing gene expression data is concentrated to a few “characteristic modes” or “eigengenes”, and why previously inferred models have a large number of false positive and false negative links. An interampatte system is characterized by strong INTERactions enabling simultaneous AMPlification and ATTEnuation of different signals and we show that perturbation of individual state variables, e.g. genes, typically leads to ill-conditioned data with both characteristic and weak modes. The weak modes are typically dominated by measurement noise due to poor excitation and their existence hampers network reconstruction. The excitation problem is solved by iterative design of correlated multi-gene perturbation experiments that counteract the intrinsic signal attenuation of the system. The next perturbation should be designed such that the expected response practically spans an additional dimension of the state space. The proposed design is numerically demonstrated for the Snf1 signalling pathway in S. cerevisiae. The impact of unperturbed and unobserved latent state variables, that exist in any real biological system, on the inferred network and required set-up of the experiments for network inference is analysed. Their existence implies that a subnetwork of pseudo-direct causal regulatory influences, accounting for all environmental effects, in general is inferred. In principle, the number of latent states and different paths between the nodes of the network can be estimated, but their identity cannot be determined unless they are observed or perturbed directly. Network inference is recognized as a variable/model selection problem and solved by considering all possible models of a specified class that can explain the data at a desired significance level, and by classifying only the links present in all of these models as existing. As shown, these links can be determined without any parameter estimation by reformulating the variable selection problem as a robust rank problem. Solution of the rank problem enable assignment of confidence to individual interactions, without resorting to any approximation or asymptotic results. This is demonstrated by reverse engineering of the synthetic IRMA gene regulatory network from published data. A previously unknown activation of transcription of SWI5 by CBF1 in the IRMA strain of S. cerevisiae is proven to exist, which serves to illustrate that even the accumulated knowledge of well studied genes is incomplete. / Denna avhandling behandlar inferens av biologiskanätverk från in vivo data genererat genom störningsexperiment, d.v.s. bestämning av kausala kopplingar som existerar mellan de observerade variablerna. Kunskap om dessa regulatoriska influenser är väsentlig för biologisk förståelse. En system egenskap—förstärksvagning—introduceras. Denna förklarar varför variationen i existerande genexpressionsdata är koncentrerat till några få ”karakteristiska moder” eller ”egengener” och varför de modeller som konstruerats innan innehåller många falska positiva och falska negativa linkar. Ett system med förstärksvagning karakteriseras av starka kopplingar som möjliggör simultan FÖRSTÄRKning och förSVAGNING av olika signaler. Vi demonstrerar att störning av individuella tillståndsvariabler, t.ex. gener, typiskt leder till illakonditionerat data med både karakteristiska och svaga moder. De svaga moderna domineras typiskt av mätbrus p.g.a. dålig excitering och försvårar rekonstruktion av nätverket. Excitationsproblemet löses med iterativdesign av experiment där korrelerade störningar i multipla gener motverkar systemets inneboende försvagning av signaller. Följande störning bör designas så att det förväntade svaret praktiskt spänner ytterligare en dimension av tillståndsrummet. Den föreslagna designen demonstreras numeriskt för Snf1 signalleringsvägen i S. cerevisiae. Påverkan av ostörda och icke observerade latenta tillståndsvariabler, som existerar i varje verkligt biologiskt system, på konstruerade nätverk och planeringen av experiment för nätverksinferens analyseras. Existens av dessa tillståndsvariabler innebär att delnätverk med pseudo-direkta regulatoriska influenser, som kompenserar för miljöeffekter, generellt bestäms. I princip så kan antalet latenta tillstånd och alternativa vägar mellan noder i nätverket bestämmas, men deras identitet kan ej bestämmas om de inte direkt observeras eller störs. Nätverksinferens behandlas som ett variabel-/modelselektionsproblem och löses genom att undersöka alla modeller inom en vald klass som kan förklara datat på den önskade signifikansnivån, samt klassificera endast linkar som är närvarande i alla dessa modeller som existerande. Dessa linkar kan bestämmas utan estimering av parametrar genom att skriva om variabelselektionsproblemet som ett robustrangproblem. Lösning av rangproblemet möjliggör att statistisk konfidens kan tillskrivas individuella linkar utan approximationer eller asymptotiska betraktningar. Detta demonstreras genom rekonstruktion av det syntetiska IRMA genreglernätverket från publicerat data. En tidigare okänd aktivering av transkription av SWI5 av CBF1 i IRMA stammen av S. cerevisiae bevisas. Detta illustrerar att t.o.m. den ackumulerade kunskapen om välstuderade gener är ofullständig. / <p>QC 20130508</p>

No description available.

Acoustic Event Detection / The detection of the Acoustic Events (AEs) naturally produced in a meeting room may help to describe the human and social activity. The automatic description of interactions between humans and environment can be useful for providing: implicit assistance to the people inside the room, context-aware and content-aware information requiring a minimum of human attention or interruptions, support for high-level analysis of the underlying acoustic scene, etc. On the other hand, the recent fast growth of available audio or audiovisual content strongly demands tools for analyzing, indexing, searching and retrieving the available documents. Given an audio document, the first processing step usually is audio segmentation (AS), i.e. the partitioning of the input audio stream into acoustically homogeneous regions which are labelled according to a predefined broad set of classes like speech, music, noise, etc. Acoustic event detection (AED) is the objective of this thesis work. A variety of features coming not only from audio but also from the video modality is proposed to deal with that detection problem in meeting-room and broadcast news domains. Two basic detection approaches are investigated in this work: a joint segmentation and classification using Hidden Markov Models (HMMs) with Gaussian Mixture Densities (GMMs), and a detection-by-classification approach using discriminative Support Vector Machines (SVMs). For the first case, a fast one-pass-training feature selection algorithm is developed in this thesis to select, for each AE class, the subset of multimodal features that shows the best detection rate. AED in meeting-room environments aims at processing the signals collected by distant microphones and video cameras in order to obtain the temporal sequence of (possibly overlapped) AEs that have been produced in the room. When applied to interactive seminars with a certain degree of spontaneity, the detection of acoustic events from only the audio modality alone shows a large amount of errors, which is mostly due to the temporal overlaps of sounds. This thesis includes several novelties regarding the task of multimodal AED. Firstly, the use of video features. Since in the video modality the acoustic sources do not overlap (except for occlusions), the proposed features improve AED in such rather spontaneous scenario recordings. Secondly, the inclusion of acoustic localization features, which, in combination with the usual spectro-temporal audio features, yield a further improvement in recognition rate. Thirdly, the comparison of feature-level and decision-level fusion strategies for the combination of audio and video modalities. In the later case, the system output scores are combined using two statistical approaches: weighted arithmetical mean and fuzzy integral. On the other hand, due to the scarcity of annotated multimodal data, and, in particular, of data with temporal sound overlaps, a new multimodal database with a rich variety of meeting-room AEs has been recorded and manually annotated, and it has been made publicly available for research purposes.

This thesis analyzes the human action recognition problem. Human actions are modeled as a time evolving temporal texture. Gabor filters, which are proved to be a robust 2D texture representation tool by detecting spatial points with high variation, is extended to 3D domain to capture motion texture features. A well known filtering algorithm and a recent unsupervised clustering algorithm, the Genetic Chromodynamics, are combined to select salient spatio-temporal features of the temporal texture and to segment the activity sequence into temporal texture primitives. Each activity sequence is represented as a composition of temporal texture primitives with its salient spatio-temporal features, which are also the symbols of our codebook. To overcome temporal variation between different performances of the same action, a Profile Hidden Markov Model is applied with Viterbi Path Counting (ensemble training). Not only parameters and structure but also codebook is learned during training.

This dissertation presents segmental discriminative analysis techniques for American Sign Language (ASL) recognition and verification. ASL recognition is a sequence classification problem. One of the most successful techniques for recognizing ASL is the hidden Markov model (HMM) and its variants. This dissertation addresses two problems in sign recognition by HMMs. The first is discriminative feature selection for temporally-correlated data. Temporal correlation in sequences often causes difficulties in feature selection. To mitigate this problem, this dissertation proposes segmentally-boosted HMMs (SBHMMs), which construct the state-optimized features in a segmental and discriminative manner. The second problem is the decomposition of ASL signs for efficient and accurate recognition. For this problem, this dissertation proposes discriminative state-space clustering (DISC), a data-driven method of automatically extracting sub-sign units by state-tying from the results of feature selection. DISC and SBHMMs can jointly search for discriminative feature sets and representation units of ASL recognition. ASL verification, which determines whether an input signing sequence matches a pre-defined phrase, shares similarities with ASL recognition, but it has more prior knowledge and a higher expectation of accuracy. Therefore, ASL verification requires additional discriminative analysis not only in utilizing prior knowledge but also in actively selecting a set of phrases that have a high expectation of verification accuracy in the service of improving the experience of users. This dissertation describes ASL verification using CopyCat, an ASL game that helps deaf children acquire language abilities at an early age. It then presents the "probe" technique which automatically searches for an optimal threshold for verification using prior knowledge and BIG, a bi-gram error-ranking predictor which efficiently selects/creates phrases that, based on the previous performance of existing verification systems, should have high verification accuracy. This work demonstrates the utility of the described technologies in a series of experiments. SBHMMs are validated in ASL phrase recognition as well as various other applications such as lip reading and speech recognition. DISC-SBHMMs consistently produce fewer errors than traditional HMMs and SBHMMs in recognizing ASL phrases using an instrumented glove. Probe achieves verification efficacy comparable to the optimum obtained from manually exhaustive search. Finally, when verifying phrases in CopyCat, BIG predicts which CopyCat phrases, even unseen in training, will have the best verification accuracy with results comparable to much more computationally intensive methods.

Η εξόρυξη κειμένου (text mining) είναι ένας νέος ερευνητικός τομέας που προσπαθεί να επιλύσει το πρόβλημα της υπερφόρτωσης πληροφοριών με τη χρησιμοποίηση των τεχνικών από την εξόρυξη από δεδομένα (data mining), την μηχανική μάθηση (machine learning), την επεξεργασία φυσικής γλώσσας (natural language processing), την ανάκτηση πληροφορίας (information retrieval), την εξαγωγή πληροφορίας (information extraction) και τη διαχείριση γνώσης (knowledge management). Στο πρώτο μέρος αυτής της διπλωματικής εργασίας αναφερόμαστε αναλυτικά στον καινούριο αυτό ερευνητικό τομέα, διαχωρίζοντάς τον από άλλους παρεμφερείς τομείς. Ο κύριος στόχος του text mining είναι να βοηθήσει τους χρήστες να εξαγάγουν πληροφορίες από μεγάλους κειμενικούς πόρους. Δύο από τους σημαντικότερους στόχους είναι η κατηγοριοποίηση και η ομαδοποίηση εγγράφων. Υπάρχει μια αυξανόμενη ανησυχία για την ομαδοποίηση κειμένων λόγω της εκρηκτικής αύξησης του WWW, των ψηφιακών βιβλιοθηκών, των ιατρικών δεδομένων, κ.λ.π.. Τα κρισιμότερα προβλήματα για την ομαδοποίηση εγγράφων είναι η υψηλή διαστατικότητα του κειμένου φυσικής γλώσσας και η επιλογή των χαρακτηριστικών γνωρισμάτων που χρησιμοποιούνται για να αντιπροσωπεύσουν μια περιοχή. Κατά συνέπεια, ένας αυξανόμενος αριθμός ερευνητών έχει επικεντρωθεί στην έρευνα για τη σχετική αποτελεσματικότητα των διάφορων τεχνικών μείωσης διάστασης και της σχέσης μεταξύ των επιλεγμένων χαρακτηριστικών γνωρισμάτων που χρησιμοποιούνται για να αντιπροσωπεύσουν το κείμενο και την ποιότητα της τελικής ομαδοποίησης. Υπάρχουν δύο σημαντικοί τύποι τεχνικών μείωσης διάστασης: οι μέθοδοι «μετασχηματισμού» και οι μέθοδοι «επιλογής». Στο δεύτερο μέρος αυτής τη διπλωματικής εργασίας, παρουσιάζουμε μια καινούρια μέθοδο «επιλογής» που προσπαθεί να αντιμετωπίσει αυτά τα προβλήματα. Η προτεινόμενη μεθοδολογία είναι βασισμένη στους κανόνες συσχέτισης (Association Rule Mining). Παρουσιάζουμε επίσης και αναλύουμε τις εμπειρικές δοκιμές, οι οποίες καταδεικνύουν την απόδοση της προτεινόμενης μεθοδολογίας. Μέσα από τα αποτελέσματα που λάβαμε διαπιστώσαμε ότι η διάσταση μειώθηκε. Όσο όμως προσπαθούσαμε, βάσει της μεθοδολογίας μας, να την μειώσουμε περισσότερο τόσο χανόταν η ακρίβεια στα αποτελέσματα. Έγινε μια προσπάθεια βελτίωσης των αποτελεσμάτων μέσα από μια διαφορετική επιλογή των χαρακτηριστικών γνωρισμάτων. Τέτοιες προσπάθειες συνεχίζονται και σήμερα. Σημαντική επίσης στην ομαδοποίηση των κειμένων είναι και η επιλογή του μέτρου ομοιότητας. Στην παρούσα διπλωματική αναφέρουμε διάφορα τέτοια μέτρα που υπάρχουν στην βιβλιογραφία, ενώ σε σχετική εφαρμογή κάνουμε σύγκριση αυτών. Η εργασία συνολικά αποτελείται από 7 κεφάλαια: Στο πρώτο κεφάλαιο γίνεται μια σύντομη ανασκόπηση σχετικά με το text mining. Στο δεύτερο κεφάλαιο περιγράφονται οι στόχοι, οι μέθοδοι και τα εργαλεία που χρησιμοποιεί η εξόρυξη κειμένου. Στο τρίτο κεφάλαιο παρουσιάζεται ο τρόπος αναπαράστασης των κειμένων, τα διάφορα μέτρα ομοιότητας καθώς και μια εφαρμογή σύγκρισης αυτών. Στο τέταρτο κεφάλαιο αναφέρουμε τις διάφορες μεθόδους μείωσης της διάστασης και στο πέμπτο παρουσιάζουμε την δικιά μας μεθοδολογία για το πρόβλημα. Έπειτα στο έκτο κεφάλαιο εφαρμόζουμε την μεθοδολογία μας σε πειραματικά δεδομένα. Η εργασία κλείνει με τα συμπεράσματα μας και κατευθύνσεις για μελλοντική έρευνα. / Text mining is a new searching field which tries to solve the problem of information overloading by using techniques from data mining, natural language processing, information retrieval, information extraction and knowledge management. At the first part of this diplomatic paper we detailed refer to this new searching field, separated it from all the others relative fields. The main target of text mining is helping users to extract information from big text resources. Two of the most important tasks are document categorization and document clustering. There is an increasing concern in document clustering due to explosive growth of the WWW, digital libraries, technical documentation, medical data, etc. The most critical problems for document clustering are the high dimensionality of the natural language text and the choice of features used to represent a domain. Thus, an increasing number of researchers have concentrated on the investigation of the relative effectiveness of various dimension reduction techniques and of the relationship between the selected features used to represent text and the quality of the final clustering. There are two important types of techniques that reduce dimension: transformation methods and selection methods. At the second part of this diplomatic paper we represent a new selection method trying to tackle these problems. The proposed methodology is based on Association Rule Mining. We also present and analyze empirical tests, which demonstrate the performance of the proposed methodology. Through the results that we obtained we found out that dimension has been reduced. However, the more we have been trying to reduce it, according to methodology, the bigger loss of precision we have been taking. There has been an effort for improving the results through a different feature selection. That kind of efforts are taking place even today. In document clustering is also important the choice of the similarity measure. In this diplomatic paper we refer several of these measures that exist to bibliography and we compare them in relative application. The paper totally has seven chapters. At the first chapter there is a brief review about text mining. At the second chapter we describe the tasks, the methods and the tools are used in text mining. At the third chapter we give the way of document representation, the various similarity measures and an application to compare them. At the fourth chapter we refer different kind of methods that reduce dimensions and at the fifth chapter we represent our own methodology for the problem. After that at the sixth chapter we apply our methodology to experimental data. The paper ends up with our conclusions and directions for future research.