by Kuo Yin-Hung. / Thesis (M.Phil.)--Chinese University of Hong Kong, 2001. / Includes bibliographical references (leaves 91-97). / Abstracts in English and Chinese. / Abstract --- p.ii / Acknowledgments --- p.iv / Chapter 1 --- Introduction --- p.1 / Chapter 1.1 --- Web Document Classification --- p.3 / Chapter 1.2 --- Web Document Integration --- p.5 / Chapter 1.3 --- Dictionary and Incremental Update --- p.5 / Chapter 1.4 --- IR-Tree --- p.6 / Chapter 1.5 --- Thesis Overview --- p.7 / Chapter 2 --- Related Works --- p.9 / Chapter 2.1 --- Semi-structured Data and OEM --- p.9 / Chapter 2.1.1 --- Semi-structured Data --- p.9 / Chapter 2.1.2 --- Object Exchange Model --- p.10 / Chapter 2.2 --- Related Work on Web Document Partitioning --- p.11 / Chapter 2.2.1 --- Retrieval of Authoritatives --- p.12 / Chapter 2.2.2 --- Document Categorization Methodology --- p.13 / Chapter 2.3 --- Semi-structured Data Indexing --- p.14 / Chapter 2.3.1 --- Lore --- p.14 / Chapter 2.3.2 --- Tsimmis --- p.15 / Chapter 2.3.3 --- Other Algorithms --- p.15 / Chapter 2.4 --- Related Work on SAMs --- p.15 / Chapter 2.4.1 --- R-Tree and R*-Tree --- p.16 / Chapter 2.4.2 --- SS-Tree and SR-Tree --- p.16 / Chapter 2.4.3 --- TV-Tree and X-Tree --- p.18 / Chapter 2.5 --- Clustering Algorithms --- p.18 / Chapter 2.5.1 --- DBSCAN and Incremental-DBSCAN --- p.20 / Chapter 3 --- Web Document Classification --- p.21 / Chapter 3.1 --- Basic Definitions --- p.21 / Chapter 3.2 --- Similarity Computation --- p.26 / Chapter 3.2.1 --- Structural Transformation --- p.27 / Chapter 3.2.2 --- Node Similarity --- p.29 / Chapter 3.2.3 --- Edge Label Similarity --- p.30 / Chapter 3.2.4 --- Structural Similarity --- p.31 / Chapter 3.2.5 --- Overall Similarity --- p.32 / Chapter 3.2.6 --- Representative Selection --- p.33 / Chapter 3.3 --- Incremental Update --- p.34 / Chapter 3.3.1 --- Documents related to a subset --- p.35 / Chapter 3.3.2 --- Documents unrelated to any subset --- p.35 / Chapter 3.3.3 --- Documents linking up two or more subsets --- p.35 / Chapter 3.4 --- Experimental Results --- p.36 / Chapter 3.4.1 --- Compare with K-NN --- p.36 / Chapter 3.4.2 --- Representative vs Feature Vector --- p.38 / Chapter 4 --- Web Document Integration --- p.40 / Chapter 4.1 --- Structure Borrowing --- p.40 / Chapter 4.2 --- Integration of Seeds --- p.42 / Chapter 4.3 --- Incremental Update --- p.48 / Chapter 4.3.1 --- New OEM record is a normal record --- p.49 / Chapter 4.3.2 --- New record is a potential seed --- p.50 / Chapter 5 --- Dictionary --- p.51 / Chapter 5.1 --- Structure of a Dictionary Entry --- p.52 / Chapter 5.2 --- Dictionary: Relation Identifier --- p.54 / Chapter 5.3 --- Dictionary: Complement of Representative --- p.55 / Chapter 5.4 --- Incremental Update --- p.56 / Chapter 5.5 --- Experimental Result --- p.57 / Chapter 5.5.1 --- Search based on keyword --- p.57 / Chapter 5.5.2 --- Search by submitting ambiguous words --- p.58 / Chapter 5.5.3 --- Retrieval of related words --- p.59 / Chapter 6 --- Structured Data Manipulation: IR-Tree --- p.61 / Chapter 6.1 --- Range Search vs Nearest Neighbor Search --- p.61 / Chapter 6.2 --- Why R*-Tree and Incremental-DBSCAN? --- p.63 / Chapter 6.3 --- IR-Tree: The Integration of Clustering and Indexing --- p.64 / Chapter 6.3.1 --- Index Structure --- p.64 / Chapter 6.3.2 --- Insertion of IR-Tree --- p.66 / Chapter 6.3.3 --- Deletion on IR-tree --- p.68 / Chapter 6.3.4 --- Nearest Neighbor Search --- p.69 / Chapter 6.3.5 --- Discussion on IR-Tree --- p.73 / Chapter 6.4 --- Experimental Results --- p.73 / Chapter 6.4.1 --- General knn-search performance --- p.74 / Chapter 6.4.2 --- Performance on Varying Dimensionality and Distribution --- p.76 / Chapter 7 --- IM-Tree: An Review --- p.80 / Chapter 7.1 --- Indexing Techniques on Metric Space --- p.80 / Chapter 7.1.1 --- Definition --- p.81 / Chapter 7.1.2 --- Metric Space Indexing Algorithms --- p.81 / Chapter 7.2 --- Clustering Algorithms on Metric Space --- p.83 / Chapter 7.3 --- The Integration of Clustering and Metric-Space Indexing Algorithm --- p.84 / Chapter 7.4 --- Proposed Algorithm --- p.85 / Chapter 7.4.1 --- Index Structure --- p.85 / Chapter 7.4.2 --- Nearest Neighbor Search --- p.86 / Chapter 7.5 --- Future Works --- p.86 / Chapter 8 --- Conclusion and Future Works --- p.87 / Chapter 8.1 --- Semi-structured Data Manipulation --- p.88 / Chapter 8.2 --- Structured Data Manipulation --- p.89

We study the problem of updating XML views defined over XML documents. A view update is performed by finding the base updates over the underlying data sources that achieve the desired view update. If such base updates do not exist, the view update is said to be untranslatable and rejected. In SQL, determining whether a view update is translatable is performed using schema level analysis, where the view definition and the base schema are used. XML schemas are more complex than SQL schemas, and can specify recursive types and cardinality constraints. There are two kinds of view updates: single view element update, where the user requires for an update over a particular view element, and a set of view elements update, where the user requires for an update over all view elements that satisfy a given XPath over the view. Accordingly, we propose one solution for each kind of view update problems based on schema level analysis for determining whether an update over XML views is translatable and for finding the translation if one exists, while considering the features of XML schemas.

The work presented in this thesis consists of three main topics: document segmentation and classification into text and score, efficient computation with binary representations, and deep learning architectures for polyphonic music transcription and classification. In the case of musical documents, an important problem is separating text from musical score by detecting the corresponding boundary boxes. A new algorithm is proposed for pixel-wise classification of digital documents in musical score and text. It is based on a bag-of-visual-words approach and random forest classification. A robust technique for identifying bounding boxes of text and music score from the pixel-wise classification is also proposed. For efficient processing of learned models, we turn our attention to binary representations. When dealing with binary data, the use of bit-packing and bit-wise computation can reduce computational time and memory requirements considerably. Efficiency is a key factor when processing large scale datasets and in industrial applications. SPmat is an optimized framework for binary image processing. We propose a bit-packed representation for binary images that encodes both pixels and square neighborhoods, and design SPmat, an optimized framework for binary image processing, around it. Bit-packing and bit-wise computation can also be used for efficient forward propagation in deep neural networks. Quantified deep neural networks have recently been proposed with the goal of improving computational time performance and memory requirements while maintaining as much as possible classification performance. A particular type of quantized neural networks are binary neural networks in which the weights and activations are constrained to $-1$ and $+1$. In this thesis, we describe and evaluate Espresso, a novel optimized framework for fast inference of binary neural networks that takes advantage of bit-packing and bit-wise computations. Espresso is self contained, written in C/CUDA and provides optimized implementations of all the building blocks needed to perform forward propagation. Following the recent success, we further investigate Deep neural networks. They have achieved state-of-the-art results and outperformed traditional machine learning methods in many applications such as: computer vision, speech recognition, and machine translation. However, in the case of music information retrieval (MIR) and audio analysis, shallow neural networks are commonly used. The effectiveness of deep and very deep architectures for MIR and audio tasks has not been explored in detail. It is also not clear what is the best input representation for a particular task. We therefore investigate deep neural networks for the following audio analysis tasks: polyphonic music transcription, musical genre classification, and urban sound classification. We analyze the performance of common classification network architectures using different input representations, paying specific attention to residual networks. We also evaluate the robustness of these models in case of degraded audio using different combinations of training/testing data. Through experimental evaluation we show that residual networks provide consistent performance improvements when analyzing degraded audio across different representations and tasks. Finally, we present a convolutional architecture based on U-Net that can improve polyphonic music transcription performance of different baseline transcription networks. / Graduate

Une collection documentaire est généralement représentée comme un ensemble de documents mais cette modélisation ne permet pas de rendre compte des relations intertextuelles et du contexte d’interprétation d’un document. Le modèle documentaire classique trouve ses limites dans les domaines spécialisés où les besoins d’accès à l’information correspondent à des usages spécifiques et où les documents sont liés par de nombreux types de relations. Ce travail de thèse propose deux modèles permettant de prendre en compte cette complexité des collections documentaire dans les outils d’accès à l’information. Le premier modèle est basée sur l’analyse formelle et relationnelle de concepts, le deuxième est basée sur les technologies du web sémantique. Appliquées sur des objets documentaires ces modèles permettent de représenter et d’interroger de manière unifiée les descripteurs de contenu des documents et les relations intertextuelles qu’ils entretiennent. / A collection of documents is generally represented as a set of documents but this simple representation does not take into account cross references between documents, which often defines their context of interpretation. This standard document model is less adapted for specific professional uses in specialized domains in which documents are related by many various references and the access tools need to consider this complexity. We propose two models based onformal and relational concept analysis and on semantic web techniques. Applied on documentary objects, these two models represent and query in a unified way documents content descriptors and documents relations.

Most existing semi-supervised document clustering approaches are model-based clustering and can be treated as parametric model taking an assumption that the underlying clusters follow a certain pre-defined distribution. In our semi-supervised document clustering, each cluster is represented by a non-parametric probability distribution. Two approaches are designed for incorporating pairwise constraints in the document clustering approach. The first approach, term-to-term relationship approach (TR), uses pairwise constraints for capturing term-to-term dependence relationships. The second approach, linear combination approach (LC), combines the clustering objective function with the user-provided constraints linearly. Extensive experimental results show that our proposed framework is effective. / This thesis presents a new framework for automatically partitioning text documents taking into consideration of constraints given by users. Semi-supervised document clustering is developed based on pairwise constraints. Different from traditional semi-supervised document clustering approaches which assume pairwise constraints to be prepared by user beforehand, we develop a novel framework for automatically discovering pairwise constraints revealing the user grouping preference. Active learning approach for choosing informative document pairs is designed by measuring the amount of information that can be obtained by revealing judgments of document pairs. For this purpose, three models, namely, uncertainty model, generation error model, and term-to-term relationship model, are designed for measuring the informativeness of document pairs from different perspectives. Dependent active learning approach is developed by extending the active learning approach to avoid redundant document pair selection. Two models are investigated for estimating the likelihood that a document pair is redundant to previously selected document pairs, namely, KL divergence model and symmetric model. / Huang, Ruizhang. / Adviser: Wai Lam. / Source: Dissertation Abstracts International, Volume: 70-06, Section: B, page: 3600. / Thesis (Ph.D.)--Chinese University of Hong Kong, 2008. / Includes bibliographical references (leaves 117-123). / Electronic reproduction. Hong Kong : Chinese University of Hong Kong, [2012] System requirements: Adobe Acrobat Reader. Available via World Wide Web. / Electronic reproduction. [Ann Arbor, MI] : ProQuest Information and Learning, [200-] System requirements: Adobe Acrobat Reader. Available via World Wide Web. / Abstracts in English and Chinese. / School code: 1307.

by Qiuyue Wang. / "July 2002." / Thesis (Ph.D.)--Chinese University of Hong Kong, 2002. / Includes bibliographical references (p. 74-82). / Electronic reproduction. Hong Kong : Chinese University of Hong Kong, [2012] System requirements: Adobe Acrobat Reader. Available via World Wide Web. / Mode of access: World Wide Web.

Submitted by Setor de Tratamento da Informa??o - BC/PUCRS (tede2@pucrs.br) on 2016-06-07T17:56:02Z No. of bitstreams: 1 TES_JAIR_MARCOS_GIACOMINI_COMPLETO.pdf: 7295746 bytes, checksum: 07cb476768bafc0a7b5ef2034a9bbeb9 (MD5) / Made available in DSpace on 2016-06-07T17:56:02Z (GMT). No. of bitstreams: 1 TES_JAIR_MARCOS_GIACOMINI_COMPLETO.pdf: 7295746 bytes, checksum: 07cb476768bafc0a7b5ef2034a9bbeb9 (MD5) Previous issue date: 2016-03-29 / Coordena??o de Aperfei?oamento de Pessoal de N?vel Superior - CAPES / The subject of this thesis is the process of making a documentary, titled Bateia, directed by the own author of this academic work. The proposal for produce the film aims to promote, from a practice, the understanding of matters that orbit the documentary cinema. To guide the procedures of the documentary realization and the reflection/writing of the thesis, an open methodology was developed. Named in process, the methodology has as its basic premise the view that the path setting is not determined a priori manner; rather, it is the process experienced by the subject-documentarist-researcher who brings out the questions to be articulated in the study. The main concepts covered in the study are: distribution of the sensible, staging, risk of the real and decis?o-tomada. The discussion of these issues is strengthened by the fact that the subjects filmed in the documentary are members of a theater group ? Companhia de Teatro ?ntimo ? whose working method involves procedures and similar conceptions to this academic research, such as: relate theory and practice, the idea of risk as his perspective and principles of sharing as form to act.In order to articulate the experience of the production of Bateia with the theoretical concepts are invited authors from academy and also documentarists: on the first field, Jacques Ranci?re, David Bordwell, Fern?o Ramos and Jean-Louis Comolli; from the side of the directors, Robert Flaherty, Dziga Vertov, Frederick Wiseman and Jean Rouch. That procedure, in line with the general purpose of the thesis, seeks to intensify the relationship between the practice of making documentary and the reflection on that. The thesis points to the productivity of the relationship between the subject-of-the-camera and filmed people built in the interval space of sharings. It also proposes "decis?o-tomada" as a theoretical concept to promote the reflection on the documentary realization from the director's experience. / O objeto de reflex?o desta tese ? o processo de realiza??o de um document?rio, intitulado Bateia, dirigido pelo pr?prio autor do trabalho acad?mico. A proposta de realizar o filme tem como objetivo principal fomentar, a partir da pr?tica, a compreens?o de temas que orbitam em torno do cinema document?rio. Para orientar os procedimentos de realiza??o document?ria e de reflex?o/escrita da tese, foi desenvolvida uma metodologia aberta. Batizada in process, a metodologia tem como premissa b?sica a perspectiva de que a configura??o do caminho n?o ? determinada de forma aprior?stica; ao contr?rio, ? o processo vivenciado pelo sujeito-documentarista-pesquisador que faz emergir as quest?es a serem articuladas no estudo. Os principais conceitos abordados no estudo s?o: partilha do sens?vel, encena??o, risco do real e decis?o-tomada. A discuss?o sobre essas quest?es ? refor?ada pelo fato de que os sujeitos filmados no document?rio s?o integrantes de um grupo teatral, a Companhia de Teatro ?ntimo, cujo m?todo de trabalho envolve procedimentos e concep??es semelhantes aos desta pesquisa acad?mica, tais como: relacionar pr?tica e teoria, ter como perspectiva de a??o a ideia de risco e atuar tendo como alicerce princ?pios de partilha. Para articular a experi?ncia vivenciada na realiza??o de Bateia com os conceitos te?ricos, s?o convidados autores do meio acad?mico e tamb?m documentaristas: do primeiro campo, Jacques Ranci?re, David Bordwell, Fern?o Ramos e Jean-Louis Comolli; do lado dos realizadores, Robert Flaherty, Dziga Vertov, Frederick Wiseman e Jean Rouch. Esse procedimento, em conson?ncia com a proposta geral da tese, busca intensificar a rela??o entre a pr?tica da realiza??o document?ria e a reflex?o sobre esse fazer. A tese aponta para a produtividade da rela??o entre sujeito-da-c?mera e pessoas filmadas constru?da no espa?o intervalar das partilhas. Tamb?m prop?e a ?decis?o-tomada? como conceito te?rico para promover a reflex?o sobre a realiza??o document?ria a partir da experi?ncia do realizador.

Ma Chak Kei. / Thesis submitted in: July 2002. / Thesis (M.Phil.)--Chinese University of Hong Kong, 2003. / Includes bibliographical references (leaves 97-101). / Abstracts in English and Chinese. / ABSTRACT --- p.II / ACKNOWLEDGEMENTS --- p.VI / TABLE OF CONTENTS --- p.VII / LIST OF TABLES --- p.XI / LIST OF FIGURES --- p.XIII / Chapter CHAPTER 1. --- INTRODUCTION --- p.1 / Chapter 1.1 --- Addressing and Manipulating XML Data --- p.1 / Chapter 1.2 --- The Structured-Element Object Model (SEOM) --- p.3 / Chapter 1.3 --- Relate Research --- p.4 / Chapter 1.4 --- Contribution --- p.5 / Chapter 1.5 --- Thesis Overview --- p.6 / Chapter CHAPTER 2. --- BACKGROUND TECHNOLOGIES --- p.7 / Chapter 2.1 --- Overview of XML --- p.7 / Chapter 2.1.1. --- XML Basic Syntax --- p.8 / Chapter 2.1.2. --- Namespaces in XML --- p.8 / Chapter 2.2 --- Overview of XML Schema --- p.9 / Chapter 2.2.1. --- W3C XML Schema --- p.10 / Chapter 2.2.2 --- ", Schema Alternatives" --- p.13 / Chapter 2.3 --- Overview of XPath --- p.13 / Chapter 2.4 --- Overview of DOM --- p.15 / Chapter CHAPTER 3. --- OVERVIEW OF STRUCTURED-ELEMENT OBJECT MODEL (SEOM) --- p.18 / Chapter 3.1 --- Introduction --- p.18 / Chapter 3.2 --- Objectives --- p.20 / Chapter 3.3 --- General Concepts in SEOM --- p.21 / Chapter 3.3.1. --- Data Representation --- p.21 / Chapter 3.3.2. --- Data Binding --- p.24 / Chapter 3.3.3. --- Data Access --- p.25 / Chapter CHAPTER 4. --- SEOM DOCUMENT MODELING --- p.27 / Chapter 4.1 --- Data Modeling --- p.27 / Chapter 4.1.1. --- Simple XML Data Model --- p.28 / Chapter 4.1.2. --- SEOM Data Model --- p.32 / Chapter 4.2 --- Schema Modeling --- p.41 / Chapter 4.2.1. --- SEOM Schema --- p.42 / Chapter 4.2.2. --- Creating a Schema --- p.46 / Chapter CHAPTER 5. --- SEOM DOCUMENT PROCESSING --- p.51 / Chapter 5.1 --- SEOM Document Processing --- p.51 / Chapter 5.2 --- The Classes --- p.51 / Chapter 5.2.1. --- SEOM Document Class --- p.52 / Chapter 5.2.2. --- A bstract SElement Class --- p.55 / Chapter 5.2.3. --- Generic SElement Class --- p.56 / Chapter 5.2.4. --- Implementation SElement Classes --- p.57 / Chapter 5.3 --- XML Parsing and Data Binding --- p.59 / Chapter 5.3.1. --- Parsing Process --- p.60 / Chapter 5.4 --- Querying --- p.62 / Chapter 5.4.1. --- Query Wrapper and Result Wrapper --- p.62 / Chapter 5.4.2. --- Embedding in XPath --- p.68 / Chapter CHAPTER 6. --- AN WEB-BASED SEOM DOCUMENT QUERY SYSTEM --- p.71 / Chapter 6.1 --- Web-based SEOM Document Query System --- p.71 / Chapter 6.2 --- Client-Server Architecture --- p.71 / Chapter 6.3 --- The Server --- p.74 / Chapter 6.3.1. --- Data Loading --- p.74 / Chapter 6.3.2. --- Implemented SElement - R-Tree --- p.74 / Chapter 6.3.3. --- Network Interface --- p.80 / Chapter 6.4 --- Client Side --- p.82 / Chapter 6.4.1. --- The Interface --- p.82 / Chapter 6.4.2. --- Programmatic Controls --- p.85 / Chapter CHAPTER 7. --- EVALUATION --- p.88 / Chapter 7.1 --- Experiment with Synthetic Data --- p.88 / Chapter 7.2 --- Qualitative Comparison --- p.90 / Chapter 7.3 --- Advantages --- p.91 / Chapter 7.4 --- Disadvantages --- p.92 / Chapter 7.5 --- Means of Enhancement --- p.93 / Chapter CHAPTER 8. --- CONCLUSION --- p.94 / BIBLIOGRAPHY --- p.97

Orientador: João Batista Ernesto de Moraes / Banca: Helen de Castro Silva / Banca: Maria Cristiane Barbosa Galvão / Resumo: A pesquisa tem como tema "O percurso gerativo de sentido aplicado à análise documental de textos narrativos de ficção: perspectivas de utilização em bibliotecas universitárias", com o foco no estudo da análise textual buscando a concretização de sentido no encadeamento das figuras e no encadeamento dos temas presentes nos contos para a recuperação da informação. O pressuposto é que não existem ferramentas de análise de assunto adequadas para a análise de obras de ficção devido à ausência de procedimentos metodológicos que contribuam efetivamente para o tratamento desse tipo de documento para o estabelecimento do tema. Dessa forma, propõem-se a busca subsídios teórico-metodológicos no percurso gerativo de sentido, componente da Semântica Discursiva, para compreensão e extração do tema em obras de ficção. Os objetivos da pesquisa são: a análise, identificação e aplicação metodológica do percurso, buscando identificar os textos figurativos e temáticos, bem como a concretização do sentido do texto; o estudo dos atores que permeiam o encadeamento das figuras e dos temas, determinando as relações existentes nesse tipo de superestrutura narrativa; identificar e analisar o nível de experiência na atuação profissional dos bibliotecários na atividade de indexação de obras ficcionais; identificar e analisar as atitudes dos profissionais bibliotecários em relação à atividade de indexação com vistas a oferecer embasamento teórico para futuros estudos que efetivem uma proposta metodológica para a identificação do tema de obras ficcionais. O método escolhido para a realização da investigação é a pesquisa descritiva por permitir a observação, o registro, a análise e a correlação dos fatos. A entrevista estruturada e focalizada será utilizada como técnica de coleta de dados que permitirá a comparação e reflexão das respostas... (Resumo completo, clicar acesso eletrônico abaixo) / Abstract: The works of fiction have no tools for appropriate subject analysis due to the absence of methodological procedures that contribute effectively to the treatment of this kind of document to the establishment of the theme. Thus, it is proposed to search for subsidies in the gerativo theoretical and methodological journey in a sense, part of Discursive Semantic for understanding and extracting the theme in works of fiction. The objectives of the research are the analysis, identification and methodological application of the route, seeking to identify those figurative and thematic texts as well as the implementation of the meaning of the text, the study of the actors that permeate the sequence of figures and topics, determining the relationship in that kind of superstructure narrative, identify and analyze the level of experience in the professional performance of librarians in the activity Index of works of fiction, identifying and analyzing the attitudes of professional librarians in relation to the activity of indexing and the offering theoretical basis studies for future studies that put into effect that a methodological proposal for identifying the subject of fictional works. The chosen method to carry out the research is the descriptive search that allows the observation, the registration, examination and the correlation of facts. The structured and focused interview will be used as a technique for collecting data that will allow comparison and reflection of the answers. The thematic and figurative journey will be applied and analysed in short stories for the establishment of the theme (Phase 1). The collection of data was held in three libraries of UNESP Libraries Network in order to extract the techniques used by librarians to identify the topic of fictional documents (Stage 2). The final phase includes the transcription and analysis of results and reflection enabling guidelines for the analysis of documentary works fictional. / Mestre

Communication trough documents have become a key element in every company?s life, communication channels expanded to include not just print but fax, web and email. Marketing and market studies repeatedly show that personalized documents (delivered through these channels) can positively influence customer relationship hereby help to achieve business success. For IT companies these changes created an exciting new opportunity to develop new solutions and offer new services. There are several companies on the market of personalized document creation which offering high quality product and services. Therefore it might not be easy to make the right choice, to find the most fitting solution and the best vendor. The main objective of this thesis is to provide detailed information about personalized document creation in general and to introduce the market leading vendors and their technology.