Conversão grafema-fone para um sistema de reconhecimento de voz com suporte a grandes vocabulários para o português brasileiro

O processamento de voz tornou-se uma tecnologia cada vez mais baseada na modelagem automática de vasta quantidade de dados. Desta forma, o sucesso das pesquisas nesta área está diretamente ligado a existência de corpora de domínio público e outros recursos específicos, tal como um dicionário fonético. No Brasil, ao contrário do que acontece para a língua inglesa, por exemplo, não existe atualmente em domínio público um sistema de Reconhecimento Automático de Voz (RAV) para o Português Brasileiro com suporte a grandes vocabulários. Frente a este cenário, o trabalho tem como principal objetivo discutir esforços dentro da iniciativa FalaBrasil [1], criada pelo Laboratório de Processamento de Sinais (LaPS) da UFPA, apresentando pesquisas e softwares na área de RAV para o Português do Brasil. Mais especificamente, o presente trabalho discute a implementação de um sistema de reconhecimento de voz com suporte a grandes vocabulários para o Português do Brasil, utilizando a ferramenta HTK baseada em modelo oculto de Markov (HMM) e a criação de um módulo de conversão grafema-fone, utilizando técnicas de aprendizado de máquina.

Metodologia de auxílio na tomada de decisão na formação de conjuntos de unidades consumidoras de energia elétrica

A edição dos procedimentos de distribuição de energia elétrica (PRODIST) confirmou a análise comparativa entre empresas de distribuição de energia elétrica estabelecida pela Resolução 024/2000 como base para o estabelecimento das metas dos indicadores de continuidade DEC (Índice de Duração Equivalente de Interrupção por Consumidor) e FEC (Índice de Frequência Equivalente de Interrupção por Consumidor). O estabelecimento das metas é influenciado diretamente pela definição dos conjuntos de unidades consumidoras das empresas de distribuição de energia elétrica, portanto, é de interesse das empresas distribuidoras uma boa definição desses conjuntos. Este trabalho apresenta o desenvolvimento de uma metodologia baseada em técnicas de aprendizado de máquina que auxilie as empresas de distribuição de energia elétrica na tomada de decisão da definição de conjuntos de unidades consumidoras objetivando melhor compatibilidade das metas para os indicadores de DEC e FEC com a realidade dos conjuntos em relação às características ambientais e de infra-estrutura da área de concessão da empresa distribuidora.

Leitura na educação infantil: implicações da Teoria Histórico-Cultural

Pós-graduação em Educação - FFC

Aprendizado de máquina baseado em tensores e suas aplicacções para floresta de caminhos ótimos

Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq)

Modelagem da irradiação direta na incidência normal em Botucatu: aprendizado de máquina, estatístico e linke

Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq)

Estimativa da irradiação solar global pelo método de Angstrom-Prescott e técnicas de aprendizado de máquinas

Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (CAPES)

Algoritmos de aprendizado semi-supervisionado baseados em grafos aplicados na bioinformática

Fundação de Amparo à Pesquisa do Estado de São Paulo (FAPESP)

Investigating Smart Sampling as a population initialization method for Differential Evolution in continuous problems

Recently, researches have shown that the performance of metaheuristics can be affected by population initialization. Opposition-based Differential Evolution (ODE), Quasi-Oppositional Differential Evolution (QODE), and Uniform-Quasi-Opposition Differential Evolution (UQODE) are three state-of-the-art methods that improve the performance of the Differential Evolution algorithm based on population initialization and different search strategies. In a different approach to achieve similar results, this paper presents a technique to discover promising regions in a continuous search-space of an optimization problem. Using machine-learning techniques, the algorithm named Smart Sampling (SS) finds regions with high possibility of containing a global optimum. Next, a metaheuristic can be initialized inside each region to find that optimum. SS and DE were combined (originating the SSDE algorithm) to evaluate our approach, and experiments were conducted in the same set of benchmark functions used by ODE, QODE and UQODE authors. Results have shown that the total number of function evaluations required by DE to reach the global optimum can be significantly reduced and that the success rate improves if SS is employed first. Such results are also in consonance with results from the literature, stating the importance of an adequate starting population. Moreover, SS presents better efficacy to find initial populations of superior quality when compared to the other three algorithms that employ oppositional learning. Finally and most important, the SS performance in finding promising regions is independent of the employed metaheuristic with which SS is combined, making SS suitable to improve the performance of a large variety of optimization techniques. (C) 2012 Elsevier Inc. All rights reserved.

Incorporating label dependency into the binary relevance framework for multi-label classification

In multi-label classification, examples can be associated with multiple labels simultaneously. The task of learning from multi-label data can be addressed by methods that transform the multi-label classification problem into several single-label classification problems. The binary relevance approach is one of these methods, where the multi-label learning task is decomposed into several independent binary classification problems, one for each label in the set of labels, and the final labels for each example are determined by aggregating the predictions from all binary classifiers. However, this approach fails to consider any dependency among the labels. Aiming to accurately predict label combinations, in this paper we propose a simple approach that enables the binary classifiers to discover existing label dependency by themselves. An experimental study using decision trees, a kernel method as well as Naive Bayes as base-learning techniques shows the potential of the proposed approach to improve the multi-label classification performance.

Statistical methods for biomedical signal analysis and processing

Statistical modelling and statistical learning theory are two powerful analytical frameworks for analyzing signals and developing efficient processing and classification algorithms. In this thesis, these frameworks are applied for modelling and processing biomedical signals in two different contexts: ultrasound medical imaging systems and primate neural activity analysis and modelling. In the context of ultrasound medical imaging, two main applications are explored: deconvolution of signals measured from a ultrasonic transducer and automatic image segmentation and classification of prostate ultrasound scans. In the former application a stochastic model of the radio frequency signal measured from a ultrasonic transducer is derived. This model is then employed for developing in a statistical framework a regularized deconvolution procedure, for enhancing signal resolution. In the latter application, different statistical models are used to characterize images of prostate tissues, extracting different features. These features are then uses to segment the images in region of interests by means of an automatic procedure based on a statistical model of the extracted features. Finally, machine learning techniques are used for automatic classification of the different region of interests. In the context of neural activity signals, an example of bio-inspired dynamical network was developed to help in studies of motor-related processes in the brain of primate monkeys. The presented model aims to mimic the abstract functionality of a cell population in 7a parietal region of primate monkeys, during the execution of learned behavioural tasks.

Kernel Methods for Tree Structured Data

Machine learning comprises a series of techniques for automatic extraction of meaningful information from large collections of noisy data. In many real world applications, data is naturally represented in structured form. Since traditional methods in machine learning deal with vectorial information, they require an a priori form of preprocessing. Among all the learning techniques for dealing with structured data, kernel methods are recognized to have a strong theoretical background and to be effective approaches. They do not require an explicit vectorial representation of the data in terms of features, but rely on a measure of similarity between any pair of objects of a domain, the kernel function. Designing fast and good kernel functions is a challenging problem. In the case of tree structured data two issues become relevant: kernel for trees should not be sparse and should be fast to compute. The sparsity problem arises when, given a dataset and a kernel function, most structures of the dataset are completely dissimilar to one another. In those cases the classifier has too few information for making correct predictions on unseen data. In fact, it tends to produce a discriminating function behaving as the nearest neighbour rule. Sparsity is likely to arise for some standard tree kernel functions, such as the subtree and subset tree kernel, when they are applied to datasets with node labels belonging to a large domain. A second drawback of using tree kernels is the time complexity required both in learning and classification phases. Such a complexity can sometimes prevents the kernel application in scenarios involving large amount of data. This thesis proposes three contributions for resolving the above issues of kernel for trees. A first contribution aims at creating kernel functions which adapt to the statistical properties of the dataset, thus reducing its sparsity with respect to traditional tree kernel functions. Specifically, we propose to encode the input trees by an algorithm able to project the data onto a lower dimensional space with the property that similar structures are mapped similarly. By building kernel functions on the lower dimensional representation, we are able to perform inexact matchings between different inputs in the original space. A second contribution is the proposal of a novel kernel function based on the convolution kernel framework. Convolution kernel measures the similarity of two objects in terms of the similarities of their subparts. Most convolution kernels are based on counting the number of shared substructures, partially discarding information about their position in the original structure. The kernel function we propose is, instead, especially focused on this aspect. A third contribution is devoted at reducing the computational burden related to the calculation of a kernel function between a tree and a forest of trees, which is a typical operation in the classification phase and, for some algorithms, also in the learning phase. We propose a general methodology applicable to convolution kernels. Moreover, we show an instantiation of our technique when kernels such as the subtree and subset tree kernels are employed. In those cases, Direct Acyclic Graphs can be used to compactly represent shared substructures in different trees, thus reducing the computational burden and storage requirements.

Lesion detection and classification in breast cancer: evaluation of approaches based on morphological features, tracer kinetic modelling and semi-quantitative parameters in MR functional imaging (DCE-MRI)

The diagnosis, grading and classification of tumours has benefited considerably from the development of DCE-MRI which is now essential to the adequate clinical management of many tumour types due to its capability in detecting active angiogenesis. Several strategies have been proposed for DCE-MRI evaluation. Visual inspection of contrast agent concentration curves vs time is a very simple yet operator dependent procedure, therefore more objective approaches have been developed in order to facilitate comparison between studies. In so called model free approaches, descriptive or heuristic information extracted from time series raw data have been used for tissue classification. The main issue concerning these schemes is that they have not a direct interpretation in terms of physiological properties of the tissues. On the other hand, model based investigations typically involve compartmental tracer kinetic modelling and pixel-by-pixel estimation of kinetic parameters via non-linear regression applied on region of interests opportunely selected by the physician. This approach has the advantage to provide parameters directly related to the pathophysiological properties of the tissue such as vessel permeability, local regional blood flow, extraction fraction, concentration gradient between plasma and extravascular-extracellular space. Anyway, nonlinear modelling is computational demanding and the accuracy of the estimates can be affected by the signal-to-noise ratio and by the initial solutions. The principal aim of this thesis is investigate the use of semi-quantitative and quantitative parameters for segmentation and classification of breast lesion. The objectives can be subdivided as follow: describe the principal techniques to evaluate time intensity curve in DCE-MRI with focus on kinetic model proposed in literature; to evaluate the influence in parametrization choice for a classic bi-compartmental kinetic models; to evaluate the performance of a method for simultaneous tracer kinetic modelling and pixel classification; to evaluate performance of machine learning techniques training for segmentation and classification of breast lesion.

Geometrical optimization of a short-stem hip implant for the reduction of proximal stress shielding (ottimizzazione geometrica di uno stelo corto di protesi d?anca per la riduzione della schermatura del carico prossimale)

Nowadays the number of hip joints arthroplasty operations continues to increase because the elderly population is growing. Moreover, the global life expectancy is increasing and people adopt a more active way of life. For this reasons, the demand of implant revision operations is becoming more frequent. The operation procedure includes the surgical removal of the old implant and its substitution with a new one. Every time a new implant is inserted, it generates an alteration in the internal femur strain distribution, jeopardizing the remodeling process with the possibility of bone tissue loss. This is of major concern, particularly in the proximal Gruen zones, which are considered critical for implant stability and longevity. Today, different implant designs exist in the market; however there is not a clear understanding of which are the best implant design parameters to achieve mechanical optimal conditions. The aim of the study is to investigate the stress shielding effect generated by different implant design parameters on proximal femur, evaluating which ranges of those parameters lead to the most physiological conditions.

DISSECTION OF STRESS RESPONSE NETWORKS REGULATING MULTIPLE STRESSES IN RICE

Important food crops like rice are constantly exposed to various stresses that can have devastating effect on their survival and productivity. Being sessile, these highly evolved organisms have developed elaborate molecular machineries to sense a mixture of stress signals and elicit a precise response to minimize the damage. However, recent discoveries revealed that the interplay of these stress regulatory and signaling molecules is highly complex and remains largely unknown. In this work, we conducted large scale analysis of differential gene expression using advanced computational methods to dissect regulation of stress response which is at the heart of all molecular changes leading to the observed phenotypic susceptibility. One of the most important stress conditions in terms of loss of productivity is drought. We performed genomic and proteomic analysis of epigenetic and miRNA mechanisms in regulation of drought responsive genes in rice and found subsets of genes with striking properties. Overexpressed genesets included higher number of epigenetic marks, miRNA targets and transcription factors which regulate drought tolerance. On the other hand, underexpressed genesets were poor in above features but were rich in number of metabolic genes with multiple co-expression partners contributing majorly towards drought resistance. Identification and characterization of the patterns exhibited by differentially expressed genes hold key to uncover the synergistic and antagonistic components of the cross talk between stress response mechanisms. We performed meta-analysis on drought and bacterial stresses in rice and Arabidopsis, and identified hundreds of shared genes. We found high level of conservation of gene expression between these stresses. Weighted co-expression network analysis detected two tight clusters of genes made up of master transcription factors and signaling genes showing strikingly opposite expression status. To comprehensively identify the shared stress responsive genes between multiple abiotic and biotic stresses in rice, we performed meta-analyses of microarray studies from seven different abiotic and six biotic stresses separately and found more than thirteen hundred shared stress responsive genes. Various machine learning techniques utilizing these genes classified the stresses into two major classes' namely abiotic and biotic stresses and multiple classes of individual stresses with high accuracy and identified the top genes showing distinct patterns of expression. Functional enrichment and co-expression network analysis revealed the different roles of plant hormones, transcription factors in conserved and non-conserved genesets in regulation of stress response.

Tumour genomic and microenvironmental heterogeneity for integrated prediction of 5-year biochemical recurrence of prostate cancer: a retrospective cohort study.

BACKGROUND Clinical prognostic groupings for localised prostate cancers are imprecise, with 30-50% of patients recurring after image-guided radiotherapy or radical prostatectomy. We aimed to test combined genomic and microenvironmental indices in prostate cancer to improve risk stratification and complement clinical prognostic factors. METHODS We used DNA-based indices alone or in combination with intra-prostatic hypoxia measurements to develop four prognostic indices in 126 low-risk to intermediate-risk patients (Toronto cohort) who will receive image-guided radiotherapy. We validated these indices in two independent cohorts of 154 (Memorial Sloan Kettering Cancer Center cohort [MSKCC] cohort) and 117 (Cambridge cohort) radical prostatectomy specimens from low-risk to high-risk patients. We applied unsupervised and supervised machine learning techniques to the copy-number profiles of 126 pre-image-guided radiotherapy diagnostic biopsies to develop prognostic signatures. Our primary endpoint was the development of a set of prognostic measures capable of stratifying patients for risk of biochemical relapse 5 years after primary treatment. FINDINGS Biochemical relapse was associated with indices of tumour hypoxia, genomic instability, and genomic subtypes based on multivariate analyses. We identified four genomic subtypes for prostate cancer, which had different 5-year biochemical relapse-free survival. Genomic instability is prognostic for relapse in both image-guided radiotherapy (multivariate analysis hazard ratio [HR] 4·5 [95% CI 2·1-9·8]; p=0·00013; area under the receiver operator curve [AUC] 0·70 [95% CI 0·65-0·76]) and radical prostatectomy (4·0 [1·6-9·7]; p=0·0024; AUC 0·57 [0·52-0·61]) patients with prostate cancer, and its effect is magnified by intratumoral hypoxia (3·8 [1·2-12]; p=0·019; AUC 0·67 [0·61-0·73]). A novel 100-loci DNA signature accurately classified treatment outcome in the MSKCC low-risk to intermediate-risk cohort (multivariate analysis HR 6·1 [95% CI 2·0-19]; p=0·0015; AUC 0·74 [95% CI 0·65-0·83]). In the independent MSKCC and Cambridge cohorts, this signature identified low-risk to high-risk patients who were most likely to fail treatment within 18 months (combined cohorts multivariate analysis HR 2·9 [95% CI 1·4-6·0]; p=0·0039; AUC 0·68 [95% CI 0·63-0·73]), and was better at predicting biochemical relapse than 23 previously published RNA signatures. INTERPRETATION This is the first study of cancer outcome to integrate DNA-based and microenvironment-based failure indices to predict patient outcome. Patients exhibiting these aggressive features after biopsy should be entered into treatment intensification trials. FUNDING Movember Foundation, Prostate Cancer Canada, Ontario Institute for Cancer Research, Canadian Institute for Health Research, NIHR Cambridge Biomedical Research Centre, The University of Cambridge, Cancer Research UK, Cambridge Cancer Charity, Prostate Cancer UK, Hutchison Whampoa Limited, Terry Fox Research Institute, Princess Margaret Cancer Centre Foundation, PMH-Radiation Medicine Program Academic Enrichment Fund, Motorcycle Ride for Dad (Durham), Canadian Cancer Society.