The so-called Web 2.0 revolution has fundamentally changed the way people interact with the Internet. The Web has turned from a read-only infrastructure to a collaborative platform. By expressing their preferences and sharing private information, the users benefit from a personalized Web experience. Yet, these systems raise several problems in terms of \emph{privacy} and \emph{scalability}. The social platforms use the user information for commercial needs and expose the privacy and preferences of the users. Furthermore, centralized personalized systems require costly data-centers. As a consequence, existing centralized social platforms do not exploit the full extent of the personalization possibilities. In this thesis, we consider the design of social networks and social information services in the context of \emph{peer-to-peer} (P2P) networks. P2P networks are decentralized architecture, thus the users participates to the service and control their own data. This greatly improves the privacy of the users and the scalability of the system. Nevertheless, building social systems in a distributed context also comes with many challenges. The information is distributed among the users and the system has be able to efficiently locate relevant data. The contributions of this thesis are as follow. We define the \emph{cold start link prediction} problem, which consists in predicting the edges of a social network solely from the social information of the users. We propose a method based on a \emph{probabilistic graph} to solve this problem. We evaluate it on a dataset from Flickr, using the group membership as social information. Our results show that the social information indeed enables a prediction of the social network. Thus, the centralization of the information threatens the privacy of the users, hence the need for decentralized systems. We propose \textsc{SoCS}, a \emph{decentralized} algorithm for \emph{link prediction}. Recommending neighbors is a central functionality in social networks, and it is therefore crucial to propose a decentralized approach as a first step towards P2P social networks. \textsc{SoCS} relies on gossip protocols to perform a force-based embedding of the social networks. The social coordinates are then used to predict links among vertices. We show that \textsc{SoCS} is adapted to decentralized systems at it is churn resilient and has a low bandwidth consumption. We propose \textsc{GMIN}, a \emph{decentralized} platform for \emph{personalized services} based on social information. \textsc{GMIN} provides each user with neighbors that share her interests. The clustering algorithm we propose takes care to encompass all the different interests of the user, and not only the main ones. We then propose a personalized \emph{query expansion} algorithm (\textsc{GQE}) that leverages the \textsc{GMIN} neighbors. For each query, the system computes a tag centrality based on the relations between tags as seen by the user and her neighbors.

Collaborative tagging is a simple and effective method for organizing and sharing web resources using human created metadata. It has arisen out of the need for an efficient method of personal organization, as the number of digital resources in everyday lives increases. While tagging has become a proven organization scheme through its popularity and widespread use on the Web, little is known about its implications and how it may effectively be applied in different situations. This is due to the fact that tagging has evolved through several iterations of use on social software websites, rather than through a scientific or an engineering design process. The research presented in this thesis, through investigations in the domain of e-learning, seeks to understand more about the scientific nature of collaborative tagging through a number of human subject studies. While broad in scope, touching on issues in human computer interaction, knowledge representation, Web system architecture, e-learning, metadata, and information visualization, this thesis focuses on how collaborative tagging can supplement the growing metadata requirements of e-learning. I conclude by looking at how the findings may be used in future research, through using information based in the emergent social networks of social software, to automatically adapt to the needs of individual users.

Collaborative tagging is a simple and effective method for organizing and sharing web resources using human created metadata. It has arisen out of the need for an efficient method of personal organization, as the number of digital resources in everyday lives increases. While tagging has become a proven organization scheme through its popularity and widespread use on the Web, little is known about its implications and how it may effectively be applied in different situations. This is due to the fact that tagging has evolved through several iterations of use on social software websites, rather than through a scientific or an engineering design process. The research presented in this thesis, through investigations in the domain of e-learning, seeks to understand more about the scientific nature of collaborative tagging through a number of human subject studies. While broad in scope, touching on issues in human computer interaction, knowledge representation, Web system architecture, e-learning, metadata, and information visualization, this thesis focuses on how collaborative tagging can supplement the growing metadata requirements of e-learning. I conclude by looking at how the findings may be used in future research, through using information based in the emergent social networks of social software, to automatically adapt to the needs of individual users.

Šis dokumentas nagrinėja paieškos organizavimo galimybes VIPS sistemos video įrašų archyve pasitelkiant semantines web technologijas. Semantinis žiniatinklis yra dalis pasaulinio interneto, kuriame informacija gali būti suprantama ne tik žmonėms, bet ir programiniams agentams, sukuriant galimybes surasti, pasidalinti ir intergruoti informacija adug paprasčiau. Socialinis žymėjimas taip pat gali grupuoti informaciją semantiniu būdu. Dokumento pradžioje aptariama teorija ir galimi sprendimai, vėliau pateikiamas galimas sprendimo modelis. Dokumentas pabaigiamas išvadomis ir rekomendacijomis. / The aim of this paper is to present search organization opportunities in VIPS system archive in semantic manner. The Semantic Web is an evolving extension of the World Wide Web in which web content can be expressed not only in natural language, but also in a format that can be read and used by software agents, thus permitting them to find, share and integrate information more easily. Social tagging can also group information in semantic manner. Theoretical background of specified model is delivered in the beginning sections of this thesis. As experimental solution information system was introduced. Thesis is concluded and recommendations are suggested for future.

Esta dissertação apresenta a estruturação de um sistema para indexação e visualização de depoimentos de história oral em vídeo. A partir do levantamento de um referencial teórico referente à indexação, o sistema resultou em um protótipo funcional de alta fidelidade. O conteúdo para a realização deste foi obtido pela indexação de 12 depoimentos coletados pela equipe do Museu da Pessoa durante o projeto Memórias da Vila Madalena, em São Paulo (ago/2012). Acervos de História Oral como o Museu da Pessoa, o Museu da Imagem e do Som ou o Centro de Pesquisa e Documentação de História Contemporânea do Brasil / CPDOC da Fundação Getúlio Vargas, reúnem milhares de horas de depoimentos em áudio e vídeo. De uma forma geral, esses depoimentos são longas entrevistas individuais, onde diversos assuntos são abordados; o que dificulta sua análise, síntese e consequentemente, sua recuperação. A transcrição dos depoimentos permite a realização de buscas textuais para acessar assuntos específicos nas longas entrevistas. Por isso, podemos dizer que as transcrições são a principal fonte de consulta dos pesquisadores de história oral, deixando a fonte primária (o vídeo) para um eventual segundo momento da pesquisa. A presente proposta visa ampliar a recuperação das fontes primárias a partir da indexação de segmentos de vídeo, criando pontos de acesso imediato para trechos relevantes das entrevistas. Nessa abordagem, os indexadores (termos, tags ou anotações) não são associados ao vídeo completo, mas a pontos de entrada e saída (timecodes) que definem trechos específicos no vídeo. As tags combinadas com os timecodes criam novos desafios e possibilidades para indexação e navegação através de arquivos de vídeo. O sistema aqui estruturado integra conceitos e técnicas de áreas aparentemente desconectadas: metodologias de indexação, construção de taxonomias, folksonomias, visualização de dados e design de interação são integrados em um processo unificado que vai desde a coleta e indexação dos depoimentos até sua visualização e interação. / This work presents the construction of an interface for visualizing and navigating the many narratives of oral history testimonials. Collections such as those belonging to the CPDOC/FGV, the Museu da Imagem e do Som and the Museu da Pessoa, contain thousands of hours of audio and video interviews. Each one of them covers many subjects, which complicates its analysis, synthesis, indexing, and consequently its retrieval. This proposal aims to facilitate the retrieval of primary sources (audio and video) by indexing specific excerpts of testimonies. To accomplish this, technologies and methodologies from areas such as: tagging, content analysis, text mining, thesauri construction and data visualization will be applied. Hence the need for an approach that consolidates these various project phases into one unified process in which the interdependencies of each step are clear and transparent. As case study, we will use 12 testimonials collected in late 2012 by researchers from the Museu da Pessoa. By indexing these videos, we will create an interface for navigating the interview segments, now categorized by topics.

Esta dissertação apresenta a estruturação de um sistema para indexação e visualização de depoimentos de história oral em vídeo. A partir do levantamento de um referencial teórico referente à indexação, o sistema resultou em um protótipo funcional de alta fidelidade. O conteúdo para a realização deste foi obtido pela indexação de 12 depoimentos coletados pela equipe do Museu da Pessoa durante o projeto Memórias da Vila Madalena, em São Paulo (ago/2012). Acervos de História Oral como o Museu da Pessoa, o Museu da Imagem e do Som ou o Centro de Pesquisa e Documentação de História Contemporânea do Brasil / CPDOC da Fundação Getúlio Vargas, reúnem milhares de horas de depoimentos em áudio e vídeo. De uma forma geral, esses depoimentos são longas entrevistas individuais, onde diversos assuntos são abordados; o que dificulta sua análise, síntese e consequentemente, sua recuperação. A transcrição dos depoimentos permite a realização de buscas textuais para acessar assuntos específicos nas longas entrevistas. Por isso, podemos dizer que as transcrições são a principal fonte de consulta dos pesquisadores de história oral, deixando a fonte primária (o vídeo) para um eventual segundo momento da pesquisa. A presente proposta visa ampliar a recuperação das fontes primárias a partir da indexação de segmentos de vídeo, criando pontos de acesso imediato para trechos relevantes das entrevistas. Nessa abordagem, os indexadores (termos, tags ou anotações) não são associados ao vídeo completo, mas a pontos de entrada e saída (timecodes) que definem trechos específicos no vídeo. As tags combinadas com os timecodes criam novos desafios e possibilidades para indexação e navegação através de arquivos de vídeo. O sistema aqui estruturado integra conceitos e técnicas de áreas aparentemente desconectadas: metodologias de indexação, construção de taxonomias, folksonomias, visualização de dados e design de interação são integrados em um processo unificado que vai desde a coleta e indexação dos depoimentos até sua visualização e interação. / This work presents the construction of an interface for visualizing and navigating the many narratives of oral history testimonials. Collections such as those belonging to the CPDOC/FGV, the Museu da Imagem e do Som and the Museu da Pessoa, contain thousands of hours of audio and video interviews. Each one of them covers many subjects, which complicates its analysis, synthesis, indexing, and consequently its retrieval. This proposal aims to facilitate the retrieval of primary sources (audio and video) by indexing specific excerpts of testimonies. To accomplish this, technologies and methodologies from areas such as: tagging, content analysis, text mining, thesauri construction and data visualization will be applied. Hence the need for an approach that consolidates these various project phases into one unified process in which the interdependencies of each step are clear and transparent. As case study, we will use 12 testimonials collected in late 2012 by researchers from the Museu da Pessoa. By indexing these videos, we will create an interface for navigating the interview segments, now categorized by topics.

No description available.

No description available.

While news articles for print use one News Triangle, where important information is at the top of the article, online news articles are supposed to use a series of Stacked News Triangles, due to online readers text- skimming habits[1]. To identify Stacked News Triangles presence, we analyse how 68 Danish journalists annotate 31 articles. We use keyword frequency as the measure of popularity. To explore if Named Entities influence News Triangle presence, we analyse Named Entities found in the articles and keywords.We find the presence of an overall News Triangle in 30 of 31 articles, while, for the presence of Stacked News Triangles, 14 of the 31 articles have Stacked News Triangles. For Named Entities in News Triangles we cannot see what their influences is. Nonetheless, we find difference in Named Entity Types in each category (Culture, Domestic, Economy, Sports).

La technologie Web 2.0 a donné lieu à un large éventail de plates-formes de partage de photos. Il est désormais possible d’annoter des images de manière collaborative, au moyen de mots-clés; ce qui permet une gestion et une recherche efficace de ces images. Toutefois, l’annotation manuelle est laborieuse et chronophage. Au cours des dernières années, le nombre grandissant de photos annotées accessibles sur le Web a permis d'expérimenter de nouvelles méthodes d'annotation automatique d'images. L'idée est d’identifier, dans le cas d’une photo non annotée, un ensemble d'images visuellement similaires et, a fortiori, leurs mots-clés, fournis par la communauté. Il existe actuellement un nombre considérable de photos associées à des informations de localisation, c'est-à-dire géo-localisées. Nous exploiterons, dans le cadre de cette thèse, ces informations et proposerons une nouvelle approche pour l'annotation automatique d'images géo-localisées. Notre objectif est de répondre aux principales limites des approches de l'état de l'art, particulièrement concernant la qualité des annotations produites ainsi que la rapidité du processus d'annotation. Tout d'abord, nous présenterons une méthode de collecte de données annotées à partir du Web, en se basant sur la localisation des photos et les liens sociaux entre leurs auteurs. Par la suite, nous proposerons une nouvelle approche afin de résoudre l’ambiguïté propre aux tags d’utilisateurs, le tout afin d’assurer la qualité des annotations. L'approche démontre l'efficacité de l'algorithme de recherche de caractéristiques discriminantes, dit de Laplace, dans le but d’améliorer la représentation de l'annotation. En outre, une nouvelle mesure de distance entre mots-clés sera présentée, qui étend la divergence de Jensen-Shannon en tenant compte des fluctuations statistiques. Dans le but d'identifier efficacement les images visuellement proches, la thèse étend sur deux point l'algorithme d'état de l'art en comparaison d'images, appelé SURF (Speeded-Up Robust Features). Premièrement, nous présenterons une solution pour filtrer les points-clés SURF les plus significatifs, au moyen de techniques de classification, ce qui accélère l'exécution de l'algorithme. Deuxièmement, la précision du SURF sera améliorée, grâce à une comparaison itérative des images. Nous proposerons une un modèle statistique pour classer les annotations récupérées selon leur pertinence du point de vue de l'image-cible. Ce modèle combine différents critères, il est centré sur la règle de Bayes. Enfin, l'efficacité de l'approche d'annotation ainsi que celle des contributions individuelles sera démontrée expérimentalement. / In the Web 2.0 era, platforms for sharing and collaboratively annotating images with keywords, called tags, became very popular. Tags are a powerful means for organizing and retrieving photos. However, manual tagging is time consuming. Recently, the sheer amount of user-tagged photos available on the Web encouraged researchers to explore new techniques for automatic image annotation. The idea is to annotate an unlabeled image by propagating the labels of community photos that are visually similar to it. Most recently, an ever increasing amount of community photos is also associated with location information, i.e., geotagged. In this thesis, we aim at exploiting the location context and propose an approach for automatically annotating geotagged photos. Our objective is to address the main limitations of state-of-the-art approaches in terms of the quality of the produced tags and the speed of the complete annotation process. To achieve these goals, we, first, deal with the problem of collecting images with the associated metadata from online repositories. Accordingly, we introduce a strategy for data crawling that takes advantage of location information and the social relationships among the contributors of the photos. To improve the quality of the collected user-tags, we present a method for resolving their ambiguity based on tag relatedness information. In this respect, we propose an approach for representing tags as probability distributions based on the algorithm of Laplacian Score feature selection. Furthermore, we propose a new metric for calculating the distance between tag probability distributions by extending Jensen-Shannon Divergence to account for statistical fluctuations. To efficiently identify the visual neighbors, the thesis introduces two extensions to the state-of-the-art image matching algorithm, known as Speeded Up Robust Features (SURF). To speed up the matching, we present a solution for reducing the number of compared SURF descriptors based on classification techniques, while the accuracy of SURF is improved through an efficient method for iterative image matching. Furthermore, we propose a statistical model for ranking the mined annotations according to their relevance to the target image. This is achieved by combining multi-modal information in a statistical framework based on Bayes' Rule. Finally, the effectiveness of each of mentioned contributions as well as the complete automatic annotation process are evaluated experimentally.