Automatic extraction of concepts from texts and applications

The extraction of relevant terms from texts is an extensively researched task in Text- Mining. Relevant terms have been applied in areas such as Information Retrieval or document clustering and classification. However, relevance has a rather fuzzy nature since the classification of some terms as relevant or not relevant is not consensual. For instance, while words such as "president" and "republic" are generally considered relevant by human evaluators, and words like "the" and "or" are not, terms such as "read" and "finish" gather no consensus about their semantic and informativeness. Concepts, on the other hand, have a less fuzzy nature. Therefore, instead of deciding on the relevance of a term during the extraction phase, as most extractors do, I propose to first extract, from texts, what I have called generic concepts (all concepts) and postpone the decision about relevance for downstream applications, accordingly to their needs. For instance, a keyword extractor may assume that the most relevant keywords are the most frequent concepts on the documents. Moreover, most statistical extractors are incapable of extracting single-word and multi-word expressions using the same methodology. These factors led to the development of the ConceptExtractor, a statistical and language-independent methodology which is explained in Part I of this thesis. In Part II, I will show that the automatic extraction of concepts has great applicability. For instance, for the extraction of keywords from documents, using the Tf-Idf metric only on concepts yields better results than using Tf-Idf without concepts, specially for multi-words. In addition, since concepts can be semantically related to other concepts, this allows us to build implicit document descriptors. These applications led to published work. Finally, I will present some work that, although not published yet, is briefly discussed in this document.

Learning facial-expression models with crowdsourcing

The computational power is increasing day by day. Despite that, there are some tasks that are still difficult or even impossible for a computer to perform. For example, while identifying a facial expression is easy for a human, for a computer it is an area in development. To tackle this and similar issues, crowdsourcing has grown as a way to use human computation in a large scale. Crowdsourcing is a novel approach to collect labels in a fast and cheap manner, by sourcing the labels from the crowds. However, these labels lack reliability since annotators are not guaranteed to have any expertise in the field. This fact has led to a new research area where we must create or adapt annotation models to handle these weaklylabeled data. Current techniques explore the annotators’ expertise and the task difficulty as variables that influences labels’ correction. Other specific aspects are also considered by noisy-labels analysis techniques. The main contribution of this thesis is the process to collect reliable crowdsourcing labels for a facial expressions dataset. This process consists in two steps: first, we design our crowdsourcing tasks to collect annotators labels; next, we infer the true label from the collected labels by applying state-of-art crowdsourcing algorithms. At the same time, a facial expression dataset is created, containing 40.000 images and respective labels. At the end, we publish the resulting dataset.

Construction of Lean and Green indexes to measure companies’ performance

The demand for costs and time reductions in companies’ processes, in order to increase efficiency, leads companies to seek innovative management paradigms to support their needs for growth and continuous improvement. The Lean paradigm has great relevance in companies’ need for waste reduction, particularly in manufacturing companies. On the other hand the demand of companies for waste reduction has gained a new dimension not only at the material level, but also at the environmental level with the introduction of the Green paradigm. As such, manufacturing companies have been adopting practices that reduce the impact of their activities on the environment. Although nowadays many manufacturing companies already implement waste reduction practices related to Lean and Green paradigms, many of them are unable to understand specifically if their efforts are enough for the application of these practices to be successful or even if their actual performance in implementing Lean or Green practices reflects the self-assessment that they have of themselves. Thus, besides the study of the development of Lean and Green paradigms in recent years, the present dissertation has the important objective of the construction of two indexes (the Lean Index and the Green Index) enabling the measurement of the performance of Portuguese manufacturing companies relating the implementation of Lean and Green practices. The data used to create the Lean and Green indexes where obtained from the implementation of the European Manufacturing Survey 2012 in Portugal. The survey questions related to the implementation of Lean and Green practices are used as variables in the development of the model for the two indexes. For the construction of representative expressions of Lean Index and Green Index it was applied the Factorial Analysis for assigning the variables weights and aggregation.

Proportion, (dis)Harmonies, Identities

Number, dimension, geometric constructions, and the relationship between dimensions, proportion, are expressions of harmony and rhythm and considered a source of beauty in the things of nature, like crystals, plants, and animals, but also in human artefacts, as buildings and art in general. This is an old idea, which corresponds to the Pythagorean notion that the world is a harmonious mathematical creation and that in order to attend such harmony, the things we do have to obey the same mathematical laws. Proportion (dis)Harmonies and Identities, aims to promote awareness and reflection on the importance of this issue with the dissemination of researches of various fields of knowledge.

Analysis and implementation of a multi-ratio switched capacitor DC-DC converter for a supercapacitor power supply

An energy harvesting system requires an energy storing device to store the energy retrieved from the surrounding environment. This can either be a rechargeable battery or a supercapcitor. Due to the limited lifetime of rechargeable batteries, they need to be periodically replaced. Therefore, a supercapacitor, which has ideally a limitless number of charge/discharge cycles can be used to store the energy; however, a voltage regulator is required to obtain a constant output voltage as the supercapacitor discharges. This can be implemented by a Switched-Capacitor DC-DC converter which allows a complete integration in CMOS technology, although it requires several topologies in order to obtain a high efficiency. This thesis presents the complete analysis of four different topologies in order to determine expressions that allow to design and determine the optimum input voltage ranges for each topology. To better understand the parasitic effects, the implementation of the capacitors and the non-ideal effect of the switches, in 130 nm technology, were carefully studied. With these two analysis a multi-ratio SC DC-DC converter was designed with an output power of 2 mW, maximum efficiency of 77%, and a maximum output ripple, in the steady state, of 23 mV; for an input voltage swing of 2.3 V to 0.85 V. This proposed converter has four operation states that perform the conversion ratios of 1/2, 2/3, 1/1 and 3/2 and its clock frequency is automatically adjusted to produce a stable output voltage of 1 V. These features are implemented through two distinct controller circuits that use asynchronous time machines (ASM) to dynamically adjust the clock frequency and to select the active state of the converter. All the theoretical expressions as well as the behaviour of the whole system was verified using electrical simulations.

Al Berto : Pelas imagens que o espelho te devolve

Dissertação de Mestrado apresentada ao ISPA - Instituto Universitário

Património Cultural Imaterial: MEMORIAMEDIA e-Museu - métodos, técnicas e práticas / Intagible Cultural Heritage: MEMORIAMEDIA e-Museum - methods, techniques and practices

O projeto MEMORIAMEDIA tem como objetivos o estudo, a inventariação e divulgação de manifestações do património cultural imaterial: expressões orais; práticas performativas; celebrações; o saber-fazer de artes e ofícios e as práticas e conhecimentos relacionados com a natureza e o universo. O MEMORIAMEDIA iniciou em 2006, em pleno debate nacional e internacional das questões do património cultural imaterial. Este livro cruza essas discussões teóricas, metodológicas e técnicas com a caracterização do MEMORIAMEDIA. Os resultados do projeto, organizados num inventário nacional, estão publicados no site www.memoriamedia.net, onde se encontram disponíveis para consulta e partilha. Filomena Sousa é investigadora de pós-doutoramento em antropologia (FCSH/UNL) e doutorada em sociologia (ISCTE-IUL). Membro integrado no Instituto de Estudos de Literatura e Tradição - patrimónios, artes e culturas (IELT) da FCSH/UNL e consultora da Memória Imaterial CRL – organização não-governamental autora e gestora do projeto MEMORIAMEDIA. Desenvolve investigação no âmbito das políticas e instrumentos de identificação, documentação e salvaguarda do património cultural imaterial e realizou vários documentários sobre expressões culturais.

EFFECT – Efficient finite element code

The theme of this dissertation is the finite element method applied to mechanical structures. A new finite element program is developed that, besides executing different types of structural analysis, also allows the calculation of the derivatives of structural performances using the continuum method of design sensitivities analysis, with the purpose of allowing, in combination with the mathematical programming algorithms found in the commercial software MATLAB, to solve structural optimization problems. The program is called EFFECT – Efficient Finite Element Code. The object-oriented programming paradigm and specifically the C ++ programming language are used for program development. The main objective of this dissertation is to design EFFECT so that it can constitute, in this stage of development, the foundation for a program with analysis capacities similar to other open source finite element programs. In this first stage, 6 elements are implemented for linear analysis: 2-dimensional truss (Truss2D), 3-dimensional truss (Truss3D), 2-dimensional beam (Beam2D), 3-dimensional beam (Beam3D), triangular shell element (Shell3Node) and quadrilateral shell element (Shell4Node). The shell elements combine two distinct elements, one for simulating the membrane behavior and the other to simulate the plate bending behavior. The non-linear analysis capability is also developed, combining the corotational formulation with the Newton-Raphson iterative method, but at this stage is only avaiable to solve problems modeled with Beam2D elements subject to large displacements and rotations, called nonlinear geometric problems. The design sensitivity analysis capability is implemented in two elements, Truss2D and Beam2D, where are included the procedures and the analytic expressions for calculating derivatives of displacements, stress and volume performances with respect to 5 different design variables types. Finally, a set of test examples were created to validate the accuracy and consistency of the result obtained from EFFECT, by comparing them with results published in the literature or obtained with the ANSYS commercial finite element code.

Orthogonal models, structure crossing, nesting and inference

We intend to study the algebraic structure of the simple orthogonal models to use them, through binary operations as building blocks in the construction of more complex orthogonal models. We start by presenting some matrix results considering Commutative Jordan Algebras of symmetric matrices, CJAs. Next, we use these results to study the algebraic structure of orthogonal models, obtained by crossing and nesting simpler ones. Then, we study the normal models with OBS, which can also be orthogonal models. We intend to study normal models with OBS (Orthogonal Block Structure), NOBS (Normal Orthogonal Block Structure), obtaining condition for having complete and suffcient statistics, having UMVUE, is unbiased estimators with minimal covariance matrices whatever the variance components. Lastly, see ([Pereira et al. (2014)]), we study the algebraic structure of orthogonal models, mixed models whose variance covariance matrices are all positive semi definite, linear combinations of known orthogonal pairwise orthogonal projection matrices, OPOPM, and whose least square estimators, LSE, of estimable vectors are best linear unbiased estimator, BLUE, whatever the variance components, so they are uniformly BLUE, UBLUE. From the results of the algebraic structure we will get explicit expressions for the LSE of these models.

Learning domain-specific sentiment lexicons with applications to recommender systems

Search is now going beyond looking for factual information, and people wish to search for the opinions of others to help them in their own decision-making. Sentiment expressions or opinion expressions are used by users to express their opinion and embody important pieces of information, particularly in online commerce. The main problem that the present dissertation addresses is how to model text to find meaningful words that express a sentiment. In this context, I investigate the viability of automatically generating a sentiment lexicon for opinion retrieval and sentiment classification applications. For this research objective we propose to capture sentiment words that are derived from online users’ reviews. In this approach, we tackle a major challenge in sentiment analysis which is the detection of words that express subjective preference and domain-specific sentiment words such as jargon. To this aim we present a fully generative method that automatically learns a domain-specific lexicon and is fully independent of external sources. Sentiment lexicons can be applied in a broad set of applications, however popular recommendation algorithms have somehow been disconnected from sentiment analysis. Therefore, we present a study that explores the viability of applying sentiment analysis techniques to infer ratings in a recommendation algorithm. Furthermore, entities’ reputation is intrinsically associated with sentiment words that have a positive or negative relation with those entities. Hence, is provided a study that observes the viability of using a domain-specific lexicon to compute entities reputation. Finally, a recommendation system algorithm is improved with the use of sentiment-based ratings and entities reputation.

Predição precoce de risco de Pré-eclâmpsia na doença hipertensiva e diabetes na gravidez: Papel de fatores angiogénicos

RESUMO: A pré-eclâmpsia tem elevada morbi-mortalidade materna e perinatal. A sua etiologia multi-fatorial tem sido objeto de investigação, não sendo ainda totalmente conhecida. Não se conhece também a razão da diferente suscetibilidade individual e das diferentes expressões da doença. A hipertensão crónica e a diabetes são fatores de risco reconhecidos, e o adiamento da maternidade contribui para que estas duas patologias sejam atualmente mais prevalentes entre as mulheres grávidas. Uma vez que o seu quadro fisiopatológico precede em meses o quadro clínico, tem-se investigado a possibilidade de serem encontrados marcadores precoces e indicadores de risco. Em Portugal, os estudos relativos à hipertensão na gravidez são escassos, bem como a investigação sobre fatores de risco e marcadores para a mesma. No sentido de avaliar possíveis marcadores de risco para o desenvolvimento de préeclâmpsia ou complicações hipertensivas foi colhida, para esta dissertação, uma amostra de 1215 mulheres que frequentaram a consulta de Hipertensão ou de Diabetes na gravidez de um centro terciário, entre 2004 e 2013. Optou-se pela realização de três estudos independentes, abrangendo os dois primeiros um leque temporal de 9 e de 2 anos respetivamente. O primeiro, centrado na hipertensão, pesquisou, em 521 mulheres com hipertensão na presente ou em anterior gravidez, fatores de risco capazes de influenciar a progressão para pré-eclâmpsia. O segundo, direcionado para a diabetes gestacional, considerou uma amostra de 334 grávidas, parte das quais tinha também hipertensão crónica e procurou identificar fatores que contribuíram para o aparecimento de complicações hipertensivas. O terceiro estudo, realizado em 2012 e 2013, em três coortes de grávidas com hipertensão crónica, com diabetes gestacional, e sem estas patologias - procurou avaliar no 1º trimestre o comportamento de dois marcadores placentares obtidos no 1º trimestre - proteína plasmática A associada à gravidez (PAPP-A) e o fator de crescimento placentar (PlGF) - e o seu papel, quer como bio-marcadores isolados, quer em associação aos fatores de risco encontrados nos anteriores estudos, na construção de um modelo preditivo de préeclâmpsia. No primeiro estudo, a nuliparidade, a hipertensão gestacional, a fluxometria das artérias uterinas com IP superiores ao P95 entre as 20-22 semanas e a existência de restrição de crescimento fetal, foram os fatores que contribuíram para a construção de um modelo preditivo de pré-eclâmpsia. No segundo estudo, a coexistência de diabetes e hipertensão crónica agravou o prognóstico, associando-se as complicações hipertensivas à multiparidade, obesidade, idade materna e etnia negra. No terceiro estudo verificou-se uma redução da PlGf e da PAPP-A no 1º trimestre nas duas primeiras coortes, comparativamente à coorte sem patologia; na análise separada de cada coorte, quando se verificaram complicações hipertensivas ou pré-eclâmpsia, as concentrações de PlGf e PAPP-A também foram inferiores. Contudo, na elaboração de um modelo preditivo de pré-eclâmpsia, em conjunto com marcadores encontrados, apenas a PlGf pode ser integrada no modelo preditivo, o que se verificou na coorte com hipertensão crónica. Os marcadores bioquímicos em estudo tiveram valores inferiores nas coortes com patologia hipertensiva, demonstrando uma deficiente produção destas proteínas placentares nestas situações, podendo ser importante a sua pesquisa. Contudo, neste estudo, apenas na coorte de hipertensão crónica a PlGf teve participação como fator de risco, na construção de um modelo preditivo de pré-eclâmpsia.--------------------------------------------------------------------------------------------------ABSTRACT: Preeclampsia is associated with a great maternal and perinatal morbimortality. Its multifactorial etiology has been under investigation and is still insufficiently understood. The reason why there are differences in individual susceptibility and differences in expressions of the disease is still unknown. Chronic hypertension and diabetes are known risk factors for preeclampsia and maternity delay contributes to the great prevalence of these pathologies among pregnant women. As the physiopathological signs antedate by months the clinical course of the disease, early risk factors and biological markers are object of clinical research. In Portugal, scarce clinical studies were devoted to hypertension in pregnancy and to risk factors and markers of this pathology. This dissertation inquires 1215 pregnant women who were treated for hypertension or diabetes in a tertiary care center between 2004 and 2013, in order to find risk markers for hypertensive complications or preeclampsia. We conducted three independent studies for this purpose. In the first one we investigated which risk factors could influence the progression to preeclampsia in 521 pregnant women with present or past history of hypertension. The second one was conducted to find what factors were associated to hypertensive complications, with a sample of 334 pregnant women with gestational diabetes, some also with chronic hypertension, addressing the identification of the factors contributing to hypertensive complications. The third study was conducted between 2012 and 2013 with three cohorts of pregnant women, with chronic hypertension, gestational diabetes, and in the third one, pregnant women had a low risk pregnancy. The objective of the study was to evaluate the behavior of two placental markers – PAPP-A and PlGf – obtained in the first trimester, and the role of these markers as isolated biomarkers or in association with other risk factors, in order to define a predictive model of early preeclampsia. In the first study, nuliparity, gestational hypertension, uterine arteries doppler with PI above P95 between 20-22 weeks of gestation and the presence of fetal growth restriction were the markers involved in a predictive model for preeclampsia. In the second study the cohort with the coexistence of diabetes and hypertension had registered worse result and hypertensive complications were associated to multiparity, obesity, maternal age and black ethnicity. In the third study there was a reduction of the PlGf and a PAPP-A concentration for the first trimester in the two first cohorts comparatively to the low risk cohort; the separate analysis of each cohort showed that plGf and PAPP-A concentrations were reduced when hypertensive complications appeared. However, when trying to find a preeclampsia predictive model, only plGf gave significant results for being considered in the model and this was only possible in the chronic hypertension cohort. The biochemical markers investigated in this study were reduced in the cohorts when high blood pressure complications occurred, showing a defective production of these placenta proteins, and suggesting that they should be investigated as first trimester biomarkers. Nevertheless, for this research, in the cohort of chronic hypertension only PlGf had a significant result, when multivariate analysis of all the risk factors was considered for the construction of a preeclampsia predictive model.

A calculadora gráfica na promoção da escrita matemática na aprendizagem de modelos contínuos não lineares: um estudo com alunos do 11.º ano de Matemática B

Relatório de estágio de mestrado em Ensino de Matemática no 3.º Ciclo do Ensino Básico e no Ensino Secundário

Unfuzzying fuzzy parsing

Dissertação de Mestrado em Engenharia Informática

O ensino de teatro de animação: contribuições para construção de novos saberes e fazeres acerca da cultura de ensinar e aprender teatro

Tese de Doutoramento em Ciências da Educação - Especialidade de Desenvolvimento Curricular

High strain rate constitutive modeling forhistorical structures subjected to blast loading

Doctoral Thesis Civil Engineering