No description available.

No description available.

Cognitive studies of clinical comprehension reveal that expert clinicians are distinguished by their superior ability to recognize meaningful patterns of data in clinical narratives. For example, in psychiatry, the findings of hallucinations and delusions suggest the subdiagnostic hypothesis of a psychotic episode, which in turn suggests several diagnoses, including schizophrenia. This dissertation describes the design and evaluation of a system that aims to simulate an important aspect of expert comprehension: the ability to recognize clusters of findings that support sub-diagnostic hypotheses. The broad range of content in psychiatric narrative presents a formidable barrier to achieving this goal, as it contains general concepts and descriptions of the subjective experience of psychiatric patients in addition to general medical and psychiatric concepts. Lexically driven language processing of such narrative would require the exhaustive predefinition of every concept likely to be encountered. In contrast, Latent Semantic Analysis (LSA) is a corpus-based statistical model of language that learns human-like estimates of the similarity between concepts from a text corpus. LSA is adapted to create trainable models of sub-diagnostic hypotheses, which are then used to recognize related elements in psychiatric discharge summary text. The system is evaluated against an independently annotated set of psychiatric discharge summaries. System-rater agreement approached rater-rater agreement, providing support for the practical application of vector-based models of meaning in domains with broad conceptual territory. Other applications and implications are discussed, including the presentation of a prototype user interface designed to enhance novice comprehension of psychiatric discourse.

No / Modern manufacturing involves multiple stages of complex process chains where Finite Element Analysis is frequently used as a simulation method on a discretized mesh to provide an accurate estimation of factors such as stresses, strains, and displacements. The choice of the most suitable element type and density is dependent on the individual manufacturing process or treatment applied at each stage of the process chain. To map between unalike Finite Element meshes, differing in density and/or element type, an Octree spatial index was evaluated as a solution for highly performant mapping. Compared to existing solutions, the Octree spatial index introduces parallelism within index creation and provides a strategy to perform the most complex interpolation technique, Element Shape Function, in a more computationally efficient manner.

This thesis presents novel techniques to index multiple media streams in a digi- tally captured presentation. These media streams are related by the common content in a presentation. We use relevance curves to represent these relationships. These relevance curves are generated by using a mix of text processing techniques and distance measures for sparse vocabularies. These techniques are used to automatically detect slide boundaries in a presentation. Accuracy of detecting these boundaries is evaluated as a function of word error rates.

Globoidal cam indexing mechanism (GCIM) plays an important role in automation and machining tools. With the compact structure, a GCIM is able to reach the required precision on account of high stiffness and minimized backlash. The requirement to improve the indexing accuracy for GCIMs from industry applications drives the research going on. In this dissertation, two strategies to improve the indexing accuracy of GCIMs are proposed. The first strategy is by considering the manufacturing parameters involved in the processes of machining and assembly. Analytical expressions for the turret motion and indexing accuracy of grooved GCIMs have been identified. Based on the kinematic and geometric relationships between the cam and its roller-follower turret, the effects on the output of the cam mechanism due to clearances (between the cam and roller; in the roller bearing), preload (change of the distance between input and output shafts), and the cam taper angle have been investigated. As a result, the roller alternation in the cam-turret system can be analyzed. Favorable parameters for the design, machining, and assembly can be selected to manufacture such devices with improved turret motion and indexing accuracy. Worked examples are given to demonstrate the applications of the approach. The second strategy is a technique for designing torque balancing cam (TBC) systems that are composed of spring-loaded planar cams with translating followers for GCIMs. Such a device can be attached to the input shaft of a GCIM to reduce the variation of its cam rotational speed. As a result, for high-speed applications, the intensity of residual vibrations of a GCIM can be decreased and its indexing accuracy can be improved. To approximate the required counterbalancing torque curves, nonparametric rational B-splines have been applied to synthesize the planar cam motion programs. Experimental results have also been shown in a practical and high-speed application to prove such a TBC mechanism is useful and effective.

Le cadre général de cette thèse est l'indexation sémantique et la recherche d'informations, appliquée à des documents multimédias. Plus précisément, nous nous intéressons à l'indexation sémantique des concepts dans des images et vidéos par les approches d'apprentissage actif, que nous utilisons pour construire des corpus annotés. Tout au long de cette thèse, nous avons montré que les principales difficultés de cette tâche sont souvent liées, en général, à l'fossé sémantique. En outre, elles sont liées au problème de classe-déséquilibre dans les ensembles de données à grande échelle, où les concepts sont pour la plupart rares. Pour l'annotation de corpus, l'objectif principal de l'utilisation de l'apprentissage actif est d'augmenter la performance du système en utilisant que peu d'échantillons annotés que possible, ainsi minimisant les coûts de l'annotations des données (par exemple argent et temps). Dans cette thèse, nous avons contribué à plusieurs niveaux de l'indexation multimédia et nous avons proposé trois approches qui succèdent des systèmes de l'état de l'art: i) l'approche multi-apprenant (ML) qui surmonte le problème de classe-déséquilibre dans les grandes bases de données, ii) une méthode de reclassement qui améliore l'indexation vidéo, iii) nous avons évalué la normalisation en loi de puissance et de l'APC et a montré son efficacité dans l'indexation multimédia. En outre, nous avons proposé l'approche ALML qui combine le multi-apprenant avec l'apprentissage actif, et nous avons également proposé une méthode incrémentale qui accélère l'approche proposé (ALML). En outre, nous avons proposé l'approche de nettoyage actif, qui aborde la qualité des annotations. Les méthodes proposées ont été tous validées par plusieurs expériences, qui ont été menées et évaluées sur des collections à grande échelle de l'indice de benchmark internationale bien connue, appelés TRECVID. Enfin, nous avons présenté notre système d'annotation dans le monde réel basé sur l'apprentissage actif, qui a été utilisé pour mener les annotations de l'ensemble du développement de la campagne TRECVID en 2011, et nous avons présenté notre participation à la tâche d'indexation sémantique de cette campagne, dans laquelle nous nous sommes classés à la 3ème place sur 19 participants. / The general framework of this thesis is semantic indexing and information retrieval, applied to multimedia documents. More specifically, we are interested in the semantic indexing of concepts in images and videos by the active learning approaches that we use to build annotated corpus. Throughout this thesis, we have shown that the main difficulties of this task are often related, in general, to the semantic-gap. Furthermore, they are related to the class-imbalance problem in large scale datasets, where concepts are mostly sparse. For corpus annotation, the main objective of using active learning is to increase the system performance by using as few labeled samples as possible, thereby minimizing the cost of labeling data (e.g. money and time). In this thesis, we have contributed in several levels of multimedia indexing and proposed three approaches that outperform state-of-the-art systems: i) the multi-learner approach (ML) that overcomes the class-imbalance problem in large-scale datasets, ii) a re-ranking method that improves the video indexing, iii) we have evaluated the power-law normalization and the PCA and showed its effectiveness in multimedia indexing. Furthermore, we have proposed the ALML approach that combines the multi-learner with active learning, and also proposed an incremental method that speeds up ALML approach. Moreover, we have proposed the active cleaning approach, which tackles the quality of annotations. The proposed methods were validated through several experiments, which were conducted and evaluated on large-scale collections of the well-known international benchmark, called TrecVid. Finally, we have presented our real-world annotation system based on active learning, which was used to lead the annotations of the development set of TrecVid 2011 campaign, and we have presented our participation at the semantic indexing task of the mentioned campaign, in which we were ranked at the 3rd place out of 19 participants.

Cette thèse a pour but de proposer de nouvelles méthodes d'indexation des bases de données vidéo de type archive culturelle à partir des actions humaines qu'elles contiennent. Les actions humaines, représentent un aspect important des contenus multimédia, à côté des sons, images ou de la parole. L'interrogation technique principale à laquelle nous répondons est ``Comment détecter et localiser précisément et rapidement dans une vidéo une action humaine, à partir de quelques exemples de cette même action?''. Le défi relevé par cette interrogation se trouve dans la satisfaction de ces deux critères: qualité de détection et rapidité.Nous avons traité, dans une première partie, l'adaptation des mesures de similarité aux contraintes de temps de calcul et mémoire nécessaires pour avoir un système rapide de détection d'actions. Nous avons montré qu'une approche de type "alignement de séquences" couplée avec une sélection de variables permet de répondre rapidement à des requêtes et obtient une bonne qualité des résultats. L'ajout d'un filtrage préliminaire permet d'améliorer encore les performances.Dans une seconde partie de la thèse nous avons crée une méthode d'accélération de l'étage de filtrage pour obtenir une complexité de recherche sous-linéaire dans la taille de la base. Nous nous sommes basés sur le hachage sensible à la similarité et sur une nouvelle approche à l'exploration dans l'espace de hachage, adaptée à la << requête-par-détecteur >>.Nous avons testé les méthodes proposées sur une nouvelle base annotée de vidéos de grande taille destinée à la détection et localisation d'actions humaines. Nous avons montré que nos approches donnent des résultats de bonne qualité et qu'elles peuvent passer à l'échelle. / This thesis proposes new methods for indexing video collections with varied content, such as cultural archives. We focus on human actions, which represent an important cultural aspect, together with sound, images and speech. Our main technical challenge is 'How to quickly detect and precisely localize human actions in a large video collection, when these actions are given as a query through example video clips?'. Thus, the difficulty of the task is due to criteria: quality of detection and search response time.

No description available.

Orientador: Mariângela Spotti Lopes Fujita / Banca: Maria de Fátima Gonçalves Moreira Tálamo / Banca: maria Isabel Asperti Nardi / Banca: Eduardo Wense Dias / Banca: José Augusto Chaves Guimarães / Resumo: A política de indexação deve ser representada por meio de uma filosofia que reflita os objetivos do sistema. Um dos aspectos concernentes à política de indexação diz respeito à conversão retrospectiva de dados e à compatibilidade de linguagem de indexação entre integrantes de um mesmo sistema cooperativo. Com o advento das tecnologias, foi possível pensar em cooperação entre bibliotecas e, conseqüentemente, em catalogação cooperativa. Esse novo panorama da catalogação trouxe um grande avanço para área: agilização da catalogação, diminuindo o tempo de serviço e otimizando esse processo, uma vez que há o aproveitamento de registros bibliográficos copiados de outras bases de dados. No entanto, houve outra conseqüência: o processo de identificação de assunto por meio da análise foi simplificado, reduzindo-o a uma simples operação de cópia , contemplando a forma do documento, e deixando de lado o conteúdo . Essa simplificação da prática nos remete a uma questão teórica e conceitual muito discutida, porém ainda não sedimentada, na área de Biblioteconomia: a conceituação de indexação e catalogação de assunto. A tese dessa pesquisa é considerar que as incoerências cometidas durante a indexação são resultados de desconhecimento da catalogação de forma e de conteúdo pelo bibliotecário, da ausência de política e manual de indexação para bibliotecas que serviriam como orientações ao trabalho do indexador e instrumento para sua formação em serviço, além de uma sistemática para identificação de conceitos. / Abstract: The indexing policy must be represented by means of a philosophy that reflects the system's aims. One of the aspects concerning the indexing policy is relating to the data retrospective conversion and the compatibility of indexing language among members of a same cooperative system. The appearing of the technologies made it possible to think of cooperation among libraries and, consequentely, of cooperative cataloguing. The new scene of the cataloguing brought a great advance to the field: agility of cataloguing, decreasing the time of service and optimizing this process, since it is possible the utilization of bibliographic records copied from other databases. Nevertheless, there were other consequence: the process of subject's cataloguing (indexing) was simplified reducing it to a simple operation of "copy", taking into account the document's "form", disregarding the "content". This practice simplification refers to a theoretical and high discussed conceptual matter, not yet well sedimented in Librarianship: the concept of indexing and subject cataloguing. The thesis of this research considers that the incoherences made during the indexing are results of non-discerment between descriptive cataloguing and subject cataloguing by the librarian, of the absence of policy and handbook of indexing for libraries which would be useful as guidelines to the indexer's work and instrument to his education in service, besides a systematics to identify concepts. We intend to present a theoretical methodological proposal for the working up of guidelines of indexing policy for the building of cooperative catalogs in university libraries, by means of the indexing process, from the socio-cognitive approach with librarians and users in the context of the university library, and in the conceptual perspective of the theoretical aspects that involve the indexing and the subject cataloguing. / Doutor