Presented work explores the possibilities of automatic construction and expansion of semantic networks with use of machine learning methods. The main focus is put on the feature retrieving procedure for the data set. The work presents a method of semantic relation retrieval, based on distributional hypothesis and trained on the data from Czech WordNet. We also show the first results for Czech language in this area of research. Part of the thesis is also a set of software for processing and evaluating of input data and a overview and discussion about its results on real-world data. The resulting tools can process data of amount in orders of hundreds of millions of words. The research part of the thesis used Czech morphologically and syntactically annotated data, but the methods are not language dependent.

Thesis (M.Sc.(Computer Science))--University of Pretoria, 2008. / Includes bibliographical references (leaves 99-116).

Thesis (Ph. D.)--School of Electrical and Computer Engineering, Georgia Institute of Technology, 2004. Directed by Nikil Jayant. / Vita. Includes bibliographical references (leaves 150-160).

Considerable evidence indicates that the expectations individuals hold about the effects of alcohol determine, to a degree, the amount of alcohol they drink. However, the bulk of this evidence was acquired using verbally-based measures of expectancy. The present study sought to extend the validation network by utilizing an electrophysiological measure -- the P300 component of the Event Related Potentials (ERPs) -- which is thought to index fundamental neurophysiological processes sensitive to expectancy.Previous research has demonstrated that, when presented with various outcomes of alcohol consumption, heavier drinkers endorse statements that assert positive and arousing effects of alcohol, while lighter drinkers endorse sedating and negative effects of alcohol. Given the sensitivity of the P300 to violation of subjective expectancies, it was hypothesized that P300 amplitude elicited by stimuli violating one's alcohol expectancies (e.g., statements describing sedating effects of alcohol for individuals with high positive expectancies) would be correlated with the participants' alcohol expectancies measured by traditional self-report measures.Participants were presented with statements reflecting a wide range of alcohol outcome effects, which either violated or confirmed the participant's own set of alcohol expectancies, while the ERPs evoked by these stimuli were recorded. As predicted, the P 300 amplitude elicited by negative alcohol expectancy stimuli was positively correlated with the degree of endorsement of positive/arousing expectancies on the self-report measure. That is, the higher the individual's positive/arousing expectancies, the larger the P300 elicited by stimuli asserting the negative effects of alcohol. There was no significant correlation, however, between P300 amplitude elicited by positive alcohol expectancy stimuli and the degree of endorsement of negative/sedating expectancies on the self-report measure.In sum, variations in the amplitude of the P300 were consistent with the model of Alcohol Expectancies: Namely, words imputing negative/sedating effects of alcohol elicited a large P300 in individuals with high but not low positive alcohol expectancies. By indexing the brain's electrophysiological response sensitive to expectancy violations, these findings demonstrate concordance between verbal measures of alcohol expectancies, which by their very nat ure are introspective, and a psychophysiological index of expectancy thought to operate automatically and to be independent of overt responding.

Η παρούσα διατριβή πραγματεύεται την ενσωμάτωση ενός σημασιολογικού δικτύου λημμάτων σ’ ένα σύνολο εφαρμογών Διαδικτύου για την αποτελεσματική διαχείριση του περιεχομένου του Παγκόσμιου Ιστού. Τα δίκτυα σημασιολογικά συσχετισμένων λημμάτων αποτελούν ένα είδος ηλεκτρονικών λεξικών στα οποία καταγράφεται σημασιολογική πληροφορία για τα λήμματα που περιλαμβάνουν, όπου τα τελευταία αποθηκεύονται σε μια δενδρική δομή δεδομένων. Ο τρόπος δόμησης του περιεχομένου των σημασιολογικών δικτύων παρουσιάζει αρκετές ομοιότητες με την οργάνωση που ακολουθούν οι ιστοσελίδες στον Παγκόσμιο Ιστό, με αποτέλεσμα τα σημασιολογικά δίκτυα να αποτελούν έναν σημασιολογικό πόρο άμεσα αξιοποιήσιμο από ένα πλήθος εφαρμογών Διαδικτύου που καλούνται να διαχειριστούν αποδοτικά το πλήθος των δεδομένων που διακινούνται στον Παγκόσμιο Ιστό. Μετά από επισκόπηση των τεχνικών που παρουσιάζονται στη διεθνή βιβλιογραφία για τη διαχείριση του περιεχομένου του Παγκόσμιου Ιστού, προτείνεται και υλοποιείται ένα πρότυπο μοντέλο διαχείρισης ιστοσελίδων, το οποίο κάνοντας εκτεταμένη χρήση ενός εμπλουτισμένου σημασιολογικού δικτύου λημμάτων, εντοπίζει εννοιολογικές ομοιότητες μεταξύ του περιεχομένου διαφορετικών ιστοσελίδων και με βάση αυτές επιχειρεί και κατορθώνει την αυτοματοποιημένη και αποδοτική δεικτοδότηση, κατηγοριοποίηση και ταξινόμηση του πλήθους των δεδομένων του Παγκόσμιου Ιστού. Για την επίδειξη του μοντέλου διαχείρισης ιστοσελίδων που παρουσιάζεται, υιοθετούμε το μοντέλο πλοήγησης στους θεματικούς καταλόγους του Παγκόσμιου Ιστού και καταδεικνύουμε πειραματικά τη συμβολή των σημασιολογικών δικτύων σε όλα τα στάδια της δημιουργίας θεματικών καταλόγων Διαδικτύου. Συγκεκριμένα, εξετάζεται η συνεισφορά των σημασιολογικών δικτύων: (i) στον ορισμό και εμπλουτισμό των θεματικών κατηγοριών των καταλόγων του Παγκόσμιου Ιστού, (ii) στην επεξεργασία και αποσαφήνιση του περιεχομένου των ιστοσελίδων, (iii) στον αυτόματο εμπλουτισμό των θεματικών κατηγοριών ενός δικτυακού καταλόγου, (iv) στην ταξινόμηση των ιστοσελίδων που έχουν δεικτοδοτηθεί στις αντίστοιχες θεματικές κατηγορίες ενός καταλόγου, (v) στη διαχείριση των περιεχομένων των θεματικών καταλόγων με τρόπο που να διασφαλίζει την παροχή χρήσιμων ιστοσελίδων προς τους χρήστες, και τέλος (vi) στην αναζήτηση πληροφορίας στους θεματικούς καταλόγους του Παγκόσμιου Ιστού. Η επιτυχία του προτεινόμενου μοντέλου επιβεβαιώνεται από τα αποτελέσματα ενός συνόλου πειραματικών εφαρμογών που διενεργήθηκαν στο πλαίσιο της παρούσας διατριβής, όπου καταδεικνύεται η συμβολή των σημασιολογικών δικτύων στην αποτελεσματική διαχείριση των πολυάριθμων και δυναμικά μεταβαλλόμενων ιστοσελίδων του Παγκόσμιου Ιστού. Η σπουδαιότητα του προτεινόμενου μοντέλου διαχείρισης ιστοσελίδων, έγκειται στο ότι, εκτός από αυτόνομο εργαλείο διαχείρισης και οργάνωσης ιστοσελίδων, συνιστά το πρώτο επίπεδο επεξεργασίας σε ευρύτερο πεδίο εφαρμογών, όπως είναι η εξαγωγή περιλήψεων, η εξόρυξη πληροφορίας, η θεματικά προσανατολισμένη προσκομιδή ιστοσελίδων, ο υπολογισμός του ρυθμού μεταβολής των δεδομένων του Παγκόσμιου Ιστού, η ανίχνευση ιστοσελίδων με παραποιημένο περιεχόμενο, κτλ. / This dissertation addresses the incorporation of a semantic network into a set of Web-based applications for the effective management of Web content. Semantic networks are a kind of machine readable dictionaries, which encode semantic information for the lemmas they contain, where the latter are stored in a tree structure. Semantic networks store their contents in a similar way to the organization that Web pages exhibit on the Web graph; a feature that makes semantic networks readily usable by several Web applications that aim at the efficient management of the proliferating and constantly changing Web data. After an overview of the techniques that have been employed for managing the Web content, we propose and implement a novel Web data management model, which relies on an enriched semantic network for locating semantic similarities in the context of distinct Web pages. Based on these similarities, our model attempts and successfully achieves the automatic and effective indexing, categorization and ranking of the numerous pages that are available on the Web. For demonstrating the potential of our Web data management model, we adopt the navigation model in Web thematic directories and we experimentally show the contribution of semantic networks throughout the construction of Web catalogs. More specifically, we study the contribution of semantic networks in: (i) determining and enriching the thematic categories of Web directories, (ii) processing and disambiguating the contents of Web pages, (iii) automatically improving the thematic categories of Web directories, (iv) ordering Web pages that have been assigned in the respective categories of a Web directory, (v) managing the contents of Web directories in a way that ensures the availability of useful Web data to the directories’ users, and (vi) searching for information in the contents of Web directories. The contribution of our model is certified by the experimental results that we obtained from a numerous of testing applications that we run in the framework of our study. Obtained results demonstrate the contribution of semantic networks in the effective management of the dynamically evolving Web content. The practical outcome of the research presented herein, besides offering a fully-fledge infrastructure for the efficient manipulation and organization of the Web data, it can play a key role in the development of numerous applications, such as text summarization, information extraction, topical-focused crawling, measuring the Web’s evolution, spam detection, and so forth.

Thesis (B.A.)--Haverford College, Dept. of Computer Science, 2003. / Includes bibliographical references.

Thesis (M.A.)--University of New Mexico, 2005. / "May, 2005." Includes bibliographical references (leaves 162-180). Also available online.

Thesis (Ph. D.)--University of Washington, 2005. / Vita. Includes bibliographical references (p. 145-158).

Submitted by Clebson Anjos (clebson.leandro54@gmail.com) on 2016-02-16T18:41:25Z No. of bitstreams: 1 arquivototal.pdf: 4249092 bytes, checksum: 55d1ea47091d7ea9ff4bd08be7f46c6b (MD5) / Made available in DSpace on 2016-02-16T18:41:25Z (GMT). No. of bitstreams: 1 arquivototal.pdf: 4249092 bytes, checksum: 55d1ea47091d7ea9ff4bd08be7f46c6b (MD5) Previous issue date: 2015-08-27 / Coordenação de Aperfeiçoamento de Pessoal de Nível Superior - CAPES / The scientific knowledge is not just the discovery of new facts and laws, but also in its publication. Aiming to understand the complex scientific production system analysis techniques have been applied in many digital repositories. Therefore, the social network analysis (SNA) has been a topic of interest of several studies in recent years. Social networks can be established from the relationship between people or the relationship of information that can be analyzed to assist decision making. The purpose of this thesis is to present a technique that makes it possible to characterize areas of scientific research by building semantic networks vary in time. These networks are built building on the co-occurrence of keywords in the titles of scientific papers, used to represent the evolution of research themes addressed by means of publication over time. / O conhecimento científico não se resume na descoberta de fatos e leis novas, mas também em sua publicação. Visando compreender o complexo sistema de produção científica, inúmeras técnicas de análise vêm sendo aplicadas em repositórios digitais. Assim sendo, a análise de redes sociais (ARS) tem sido um tópico de interesse de diversos estudos realizados nos últimos anos. As redes sociais podem ser estabelecidas a partir da relação entre pessoas ou da relação de informações que podem ser analisadas a fim de auxiliar a tomada de decisão. A proposta desta dissertação é apresentar uma técnica que possibilite caracterizar áreas de pesquisa científica através da construção de redes semânticas variáveis no tempo. Essas redes são construídas tomando por base a coocorrência de palavras-chave nos títulos de artigos científicos, que servem para representar a evolução dos temas de pesquisa abordados pelos meios de publicação ao longo do tempo.

Made available in DSpace on 2018-06-27T18:56:11Z (GMT). No. of bitstreams: 2 Henrique Tamiosso Machado.pdf: 2591851 bytes, checksum: b09aea9b36e9a9f1d51baa431dec4f93 (MD5) Henrique Tamiosso Machado.pdf.jpg: 3371 bytes, checksum: 9b313eea24e6b21e8cdba71fdc2e37be (MD5) Previous issue date: 2009-03-12 / Coordenação de Aperfeiçoamento de Pessoal de Nível Superior / In the bioinformatics has a great amount of biological and genetic information that serve of support for research, and each day this amount of information grows more. Diverse research made for many researchers in different areas, as molecular biology, structural biochemist, enzymology, physiology, pathology, among others, comes generating some results and information that must be stored to be used in the diverse forms. But there the problem appears, how to store, to manipulate, to visualize all these information? The bioinformatics uses the computational power to catalogue, to organize, to structuralize and to manipulate these information facilitating the use of these information that are of extreme importance for Biology. The PubMed is a service of the National Library of Medicine of United States that supplies access to more than 18 million citations for periodical scientific articles in the area of health science. In order to consider a new boarding for the search and representation of the knowledge found in the gotten result, this work presents the use of norm ISO 13250 Topic Maps for the creation of semantic nets involving the concepts found in the system of information of the PubMed, the construction of a composed architecture for a data base with deriving information of the PubMed by the National Library of Medicine of the United States (NLM), an interface for different research of the available one for the Entrez system, where it is possible to define priorities for the consultations. Also is possible to make the representation of the knowledge contained in the PubMed through semantic nets and techniques of date mining. / Na bioinformática existe uma grande quantidade de informações biológicas e genéticas que servem de suporte para pesquisas, e a cada dia essa quantidade de informações cresce ainda mais. Diversas pesquisas realizadas por inúmeros pesquisadores de diferentes áreas como biologia molecular, bioquímica estrutural, enzimologia, fisiologia, patologia, entre outras, vem gerando vários resultados e informações que devem ser armazenadas para serem utilizadas de diversas formas. Mas aí que surge o problema, como armazenar, manipular, visualizar todas essas informações? A bioinformática usa o poder computacional para catalogar, organizar, estruturar e manipular essas informações de uma forma que facilite a utilização dessas informações que são de extrema importância para a Biologia. O PubMed é um serviço da Biblioteca Nacional de Medicina dos Estados Unidos (U.S. National Library of Medicine) que fornece acesso para mais de 18 milhões de citações para artigos científicos de jornais da área de ciências da saúde. De modo a propor uma nova abordagem para a busca e representação do conhecimento encontrado no resultado obtido, este trabalho apresenta uma arquitetura de um sistema para consultas utilizando prioridades e visualização das informações através de redes semânticas para representar o conhecimento contido no PubMed, para isso, foi utilizado a norma ISO 13250 Topic Maps para a criação de redes semânticas envolvendo os conceitos encontrados no sistema de informação do PubMed, realizado o desenvolvimento de uma arquitetura composta por um banco de dados com informações oriundas do PubMed disponibilizadas pela Biblioteca Nacional de Medicina dos Estados Unidos (NLM), uma interface para pesquisa diferente da disponibilizada pelo sistema Entrez, onde seja possível definir prioridades para as consultas. E também fazer a representação do conhecimento contido no PubMed através de redes semânticas e técnicas de data mining.