No description available.

Nous abordons, dans cette thèse, les méthodes de combinaison de systèmesde transcription de la parole à Large Vocabulaire. Notre étude se concentre surl’attelage de systèmes de transcription hétérogènes dans l’objectif d’améliorerla qualité de la transcription à latence contrainte. Les systèmes statistiquessont affectés par les nombreuses variabilités qui caractérisent le signal dela parole. Un seul système n’est généralement pas capable de modéliserl’ensemble de ces variabilités. La combinaison de différents systèmes detranscription repose sur l’idée d’exploiter les points forts de chacun pourobtenir une transcription finale améliorée. Les méthodes de combinaisonproposées dans la littérature sont majoritairement appliquées a posteriori,dans une architecture de transcription multi-passes. Cela nécessite un tempsde latence considérable induit par le temps d’attente requis avant l’applicationde la combinaison.Récemment, une méthode de combinaison intégrée a été proposée. Cetteméthode est basée sur le paradigme de décodage guidé (DDA :Driven DecodingAlgorithm) qui permet de combiner différents systèmes durant le décodage. Laméthode consiste à intégrer des informations en provenance de plusieurs systèmes dits auxiliaires dans le processus de décodage d’un système dit primaire.Notre contribution dans le cadre de cette thèse porte sur un double aspect : d’une part, nous proposons une étude sur la robustesse de la combinaison par décodage guidé. Nous proposons ensuite, une amélioration efficacement généralisable basée sur le décodage guidé par sac de n-grammes,appelé BONG. D’autre part, nous proposons un cadre permettant l’attelagede plusieurs systèmes mono-passe pour la construction collaborative, à latenceréduite, de la sortie de l’hypothèse de reconnaissance finale. Nous présentonsdifférents modèles théoriques de l’architecture d’attelage et nous exposons unexemple d’implémentation en utilisant une architecture client/serveur distribuée. Après la définition de l’architecture de collaboration, nous nous focalisons sur les méthodes de combinaison adaptées à la transcription automatiqueà latence réduite. Nous proposons une adaptation de la combinaison BONGpermettant la collaboration, à latence réduite, de plusieurs systèmes mono-passe fonctionnant en parallèle. Nous présentons également, une adaptationde la combinaison ROVER applicable durant le processus de décodage via unprocessus d’alignement local suivi par un processus de vote basé sur la fréquence d’apparition des mots. Les deux méthodes de combinaison proposéespermettent la réduction de la latence de la combinaison de plusieurs systèmesmono-passe avec un gain significatif du WER. / This thesis presents work in the area of Large Vocabulary ContinuousSpeech Recognition (LVCSR) system combination. The thesis focuses onmethods for harnessing heterogeneous systems in order to increase theefficiency of speech recognizer with reduced latency.Automatic Speech Recognition (ASR) is affected by many variabilitiespresent in the speech signal, therefore single ASR systems are usually unableto deal with all these variabilities. Considering these limitations, combinationmethods are proposed as alternative strategies to improve recognitionaccuracy using multiple recognizers developed at different research siteswith different recognition strategies. System combination techniques areusually used within multi-passes ASR architecture. Outputs of two or moreASR systems are combined to estimate the most likely hypothesis amongconflicting word pairs or differing hypotheses for the same part of utterance.The contribution of this thesis is twofold. First, we study and analyze theintegrated driven decoding combination method which consists in guidingthe search algorithm of a primary ASR system by the one-best hypothesesof auxiliary systems. Thus we propose some improvements in order to makethe driven decoding more efficient and generalizable. The proposed methodis called BONG and consists in using Bag Of N-Gram auxiliary hypothesisfor the driven decoding.Second, we propose a new framework for low latency paralyzed single-passspeech recognizer harnessing. We study various theoretical harnessingmodels and we present an example of harnessing implementation basedon client/server distributed architecture. Afterwards, we suggest differentcombination methods adapted to the presented harnessing architecture:first we extend the BONG combination method for low latency paralyzedsingle-pass speech recognizer systems collaboration. Then we propose, anadaptation of the ROVER combination method to be performed during thedecoding process using a local vote procedure followed by voting based onword frequencies.

No description available.

In recent years, techniques that had been developed for the combination of unification algorithms for equational theories were extended to combining constraint solvers. These techniques inherited an old deficit that was already present in the ombination of equational theories which makes them rather unsuitable for pratical use: The underlying combination algorithms are highly non-deterministic. This paper is concerned with the pratical problem of how to optimise the combination method of Baader and Schulz. We present two optimisation methods,called the iterative and the deductive method. The iterative method reorders and localises the non-deterministic decisions. The deductive method uses specific algorithms for the components to reach certain decisions deterministically. Run time tests of our implementation indicate that the optimised combination method yields combined decision procedures that are efficient enough to be used in practice.

A nominal setup at Volvo Cars Corporation is the placement determination for two adjacent exterior parts on the car. To place the parts in optimal positions, nominal values for gaps and flushes are determined. When a nominal setup becomes more complex, VSA (Vehicle System Architect) is summoned. These appearing situations regard the involvement of several attributes and the need for a combination of vari- ous contributions. There are static, dynamic (overslam or dynamic movement) and thermal contributions that are combined into nominal values of gap and flush dis- tances. The determination process of a nominal setup contains both calculation for each contribution, as well as the combination method which takes place at the VSA meetings. This Master Thesis project consists of the development of an improved determination process for nominal setups. The current determination process has a low level of transparency within the differ- ent group’s methods. Another issue is the insecurity of the probability estimations made when combining the contributions. Therefore, the focus of the project was to infuse a greater understanding of the contribution derivations, and greater insight into the probability of the taken risks. To achieve that, the project was divided into three parts; mapping of the determination process, individual contribution improve- ments and finally, improvements to the combination method. In contemplation of improving a process, plenty of knowledge needs to be gathered, regarding methods, simulations and possibilities. This was executed by interviewing experts within spe- cific areas at the different groups at VCC. Development of the improvements was done by interviews and various studies. It was shown that the mapping of the determination process increased the trans- parency between the groups as it increased the understanding of individual groups’ work. Contribution improvements lead to more realistic load cases used for dimen- sioning. A performed overslam clinic, where closing velocity data of a tailgate were collected, lead to a greater statistical base for which load case should be used. For dynamic movement, another method is proposed that considers relative movement instead of applied accelerations. For the thermal contribution, the approach of ge- ographically gathered temperature data was proposed. The improved combination method generates combinations with regard to three input values instead of one, from each contribution, to create different combination scenarios. The probabilities of the scenario occurrences are estimated which gave VCC a greater understanding of what risks that are taken. Furthermore, the combination method also educates the VSA meeting attendees by exhibiting the derivations and bases for each contri- bution.

[pt] O processo de combinação de ações de sintonia fina não possui nem uma formulação precisa, nem uma abordagem formal para solucioná-lo. É necessário definir o que combinar dentre as múltiplas ações existentes e, uma vez escolhidas, como compor de maneira que as restrições sejam verificadas. Trata-se de um problema complexo e relevante na área de bancos de dados, tanto para soluções manuais pelo DBA como automáticas, por meio de softwares especializados. Isto ocorre pois os diferentes tipos de ações de sintonia possuem estratégias distintas para alcançar o objetivo em comum. Esta tese propõe um método automático para geração e seleção de soluções combinadas de sintonia fina para bancos de dados relacionais. Discute-se como combinar soluções e respeitar as restrições tecnológicas e recursos computacionais disponíveis. Por fim, apresenta-se uma implementação e avaliação utilizando três SGBDs de mercado relevantes, em que mostramos tanto a eficácia como a eficiência do método proposto. Os resultados mostraram que o método é capaz de produzir soluções combinadas válidas mais eficientes que soluções locais independentes. / [en] The process of combining database tuning actions has neither a precise formulation nor a formal approach to solving it. It is necessary to define what to combine among multiple existing operations and, once chosen, how to compose so that constraints can be verified. It is a complex and relevant problem in the database research area, both for the DBA manual solutions, and automatic ones using specialized software. It is important because the different types of tuning actions have different strategies to achieve a common goal. This thesis proposes an automated method for generating and selecting combined tuning solutions for relational databases. It discusses how to mix solutions and still respect both the technological constraints and available computational resources. Finally, we present an implementation and evaluation using three relevant market DBMSs, where we show both the effectiveness and the efficiency of the proposed method. The results showed that the technique is capable of producing combined solutions that are more efficient than independent local solutions.