Submitted by Alice Duarte Scarpa (alicescarpa@gmail.com) on 2017-09-26T15:54:11Z No. of bitstreams: 1 dissertacao (3).pdf: 2075770 bytes, checksum: 4838890c00b2e15a62cffbbbcb4ab5a8 (MD5) / Approved for entry into archive by Janete de Oliveira Feitosa (janete.feitosa@fgv.br) on 2017-10-31T11:43:22Z (GMT) No. of bitstreams: 1 dissertacao (3).pdf: 2075770 bytes, checksum: 4838890c00b2e15a62cffbbbcb4ab5a8 (MD5) / Made available in DSpace on 2017-11-10T18:24:46Z (GMT). No. of bitstreams: 1 dissertacao (3).pdf: 2075770 bytes, checksum: 4838890c00b2e15a62cffbbbcb4ab5a8 (MD5) Previous issue date: 2017-08-24 / The vast amount of documents available nowadays presents a great opportunity for advancing Political Sciences. At the same time, this deluge of information poses a problem, because it is no longer feasible for researchers to analyze every document manually. Modern natural language processing techniques have an essential role in helping with this process. The goal of this work is to create a tool based on natural language processing techniques that helps researchers to navigate an important database, Cablegate, which is a corpus of over 250 thousand diplomatic cables sent between US embassies that was published as part of WikiLeaks. This is a very important database that can shed new light at key historical moments of the twenty-first century. / A enorme quantidade de documentos disponíveis atualmente representa um grande potencial de avanço para as Ciências Políticas. Ao mesmo tempo essa riqueza de informações gera um problema, pois não é mais possível que pesquisadores analisem todos os documentos manualmente. Técnicas modernas de processamento de linguagem natural têm um papel essencial a cumprir para auxiliar tal avanço. O objetivo desse trabalho é criar uma ferramenta baseada em processamento de linguagem de natural que ajude pesquisadores a navegar uma base de dados muito importante, o \textit{Cablegate}, que é um conjunto de mais de 250 mil cabos diplomáticos de embaixadas dos Estados Unidos que foi publicado como parte do \textit{WikiLeaks}. Essa é uma base muito importante que pode trazer uma nova luz sobre vários momentos-chave do início do século XXI.

Le présent travail s'inscrit dans le cadre d'une collaboration entre le Laboratoire de Mécanique des Fluides et d'Acoustique à l'École Centrale de Lyon (LMFA-ECL), Turbomeca et l'ONERA. Le sujet de recherche porte sur l'interaction rouet-diffuseur et sur l'entrée en régime de pompage d'un étage de compresseur centrifuge transsonique à fort taux de compression étudié à vitesse de rotation de croisière (0,927Nn). L'alimentation des analyses est réalisée par trois méthodes de mesures et deux types de simulations numériques. Le module d'essai est installé sur le banc d'essai 1 MW du LMFA. La caractérisation expérimentale du compresseur TM est réalisée par le biais de mesures de pression et température auxquelles sont adjointes des mesures du débit et de la vitesse de rotation de la roue mobile. La description de l'écoulement interne au compresseur s'appuie sur les résultats de sondages par Anémométrie Laser à effet Doppler (LDA) et de mesures de pression à haute fréquence. Les simulations numériques sont réalisées par l'intermédiaire du code de calcul elsA développé par l'ONERA, qui permet de résoudre le système d'équations de Navier-Stokes couplé à un modèle de turbulence k-l de Smith. Deux modélisations de l'interaction rouet-diffuseur permettent de générer des champs aérodynamiques stationnaires (modèle plan de mélange) et instationnaires (modèle chorochronique). La comparaison entre les données expérimentales et numériques est très satisfaisante et permet alors de profiter pleinement de la richesse des informations numériques. L'examen détaillé de l'écoulement interne au rouet pour trois points de fonctionnement (à débit bloqué, à rendement maximum et à proximité du pompage) révèle que, du blocage vers le pompage, l'évolution de l'intensité et de la taille du tourbillon de jeu est le point de départ d'un enchaînement de mécanismes conduisant à la dilatation du sillage de la structure jet-sillage. Pour l'écoulement en amont du diffuseur, ceci ce traduit en moyenne temporelle par une augmentation de l'incidence principalement au voisinage du moyeu. L'onde de choc en amont des aubes du diffuseur remonte à mesure que le débit du compresseur diminue. La trajectoire de l'écoulement principal bascule du côté de la face en dépression vers le côté de la face en pression du canal inter-aubes du diffuseur. La cartographie des nombreux décollements de couches limites est également modifiée à l'approche du pompage. Les structures instationnaires majeures sont produites par l'interaction de l'onde de choc en amont des aubes du diffuseur avec les pales du rouet. Des ondes de pression progressives et des poches à faible nombre de Mach sont ainsi générées. Les ondes pression impriment d'intenses fluctuations au champ de vitesse qui favorisent le processus de mélange. En conséquence, les couches limites sont plus robustes vis-à-vis des décollements (en moyenne temporelle). Le défilement instationnaire des poches à faible nombre de Mach engendre une dissymétrie marquée des conditions d'alimentation du diffuseur dans la direction azimutale. Au cours du changement de point de fonctionnement en allant du débit bloqué vers le pompage, les ondes de pression se renforcent et la taille des poches à faible nombre de Mach diminue. De ce fait, les conditions sont plutôt favorables à retarder l'entrée en pompage du compresseur qui est localement initié en amont du diffuseur aubé. Le pompage du compresseur est provoqué par un changement rapide de la structure supersonique de l'écoulement en entrée de diffuseur qui est alors déstabilisée par les fluctuations de pression des ondes progressives. / The present work is in line with a collaboration between the Laboratoire de Mécanique des Fluides et d'Acoustique at École Centrale de Lyon (LMFA-ECL), Turbomeca and ONERA. The subject is focused on impeller-diffuser interaction and phenomena occurring during surge ignition of a transonic centrifugal stage with high-compression ratio at cruise rotation speed (0,927Nn). The analysed data come from three measurement devices and two kinds of numerical simulations. The 1MW LMFA-ECL test rig was used for carrying out the experiments on the centrifugal compressor stage. The global performances were obtained by pressure, temperature, mass flow rate and rotation speed measurements. The internal flow field properties were probed by Laser Doppler Anemometry (LDA) and high frequency pressure measurements. The computations were performed with the elsA software developed at ONERA. The code solves the compressible Reynolds Averaged Navier-Stokes equations associated with the two-equations (k-l) turbulence model of Smith. Two models of impeller-diffuser interaction were used to simulate the flow within the compressor. The first model is based on the Averaged Passage equations, gives a stationary description of the internal flow fields. The second model is based on the phase lagged approach and allows access to the unsteady phenomena. A good agreement between experiments and simulations was obtained, which justifies the use of the CFD results for the flow field analysis. Analysis of the flow development within the impeller were carried out for three operating points (choke, peak efficiency and close to surge). This study shows that, from choke to near surge, the development of the tip clearance vortex is the starting point of a sequence of physical mechanisms which lead to an extend of the wake of the jet-wake structure. The change in flow field at impeller exit tends to increase the upstream incidence of the vaned diffuser, especially close to the hub. The vane bow shock wave progresses in the impeller passages when the operating point moves from choke to near surge. The location of the main flow moves from suction side to pressure side of the vaned diffuser passage. The topology of the boundary layer separations within the diffuser passage is deeply affected when approaching surge. The main unsteady structures are generated by the interaction between the bow shock wave and the impeller blades. Progressive pressure waves and low Mach number flow bubbles are generated by this locally time-spaced interaction phenomenon. The pressure waves lead to strong fluctuations of the velocity field. As a consequence, the boundary layer becomes more resistant in relation to the separations (in term of time-averaged point of view). In time-averaged point of view, the low Mach number flow bubbles lead to inhomogeneous conditions at the vaned diffuser inlet along the azimuthal direction . From choke to near surge operating point, the strength of the pressure wave increases whereas the size of the low Mach number flow bubbles decreases. These conditions are quite favourable to push back the surge limit which is locally initiated in the inlet part of the vaned diffuser. The surge of the compressor is produced by a rapid change in supersonic flow structure at diffuser inlet. The pressure fluctuations due to the progressive waves lead to destabilise the new supersonic flow configuration and degenerate in the global instability of the compressor flow.

Response to Intervention (RTI) has become the gateway to identification for many students with specific learning disabilities. Those include students in the 17 states that require RTI as the source of eligibility data as well as many students in the 33 states that allow districts to choose RTI among other options (e.g., discrepancy). There is concern that the use of RTI may infringe on due-process protections and child-find responsibilities. Specifically, the number of days students must remain in RTI may delay their referral for comprehensive evaluation and, hence, potential eligibility for special education. In the present study, the authors surveyed District Directors of Special Education in selected states to determine whether guidelines or policies were present to govern referrals for a comprehensive evaluation and how long a student may remain in RTI tiers before referral or eligibility determination. The extent to which districts permit individual schools to enact their own policies was also investigated. Results indicate that (a) RTI is often used to identify specific learning disabilities without clear guidelines, (b) many aspects of RTI are implemented without being communicated within state and district administrative levels, and (c) RTI is the required specific learning disability assessment determinant in a significant percentage of districts in states that allow that choice. Implications for future research and practice are discussed.

We used tweets to collect public discussion on organizations' activities during the specified Covid19 period. Through topic modeling, we were able to establish discussed topics in line with the organization's activities. Our research majored on tweets with matching hashtags W.H.O (world health organization) and coronavirus, covid19 or covid. We extracted five latent topics and explored the distribution or evolution of those topics over time. We were able to find people's opinions on hot topics (the period when a topic is mainly discussed); the hot topics reflect activities on the timeline of W.H.O during the specified period of the Pandemic. Our results show that the key topics are identified and characterized by specific events that happened during the specified period in our data. Our result describes the events that happened on the timeline of the W.H.O, showing the public opinion on each period a discussion is hot. It also shows how people's opinions revolve during the period. Our results will be helpful in identifying public sentiment on events, how people's opinion varies, and can also help understand different events of the organization based on the aim and objective of the event.

This master’s thesis is about a method for sentimental analysis, especially machine learning methods without teacher. In detail are described method for semantic modeling LSA, pLSA a LDA. It was created a LDA implementation in Java language, which was used to emotional classification of 860 Czech documents to six different emotional categories. Maximal accuracy was 24 % if optimized parameters was used.

This work deals with study of swirling flows where the spiral vortical structure appears. The main relation is to flow seen in the draft tube cone of hydraulic turbines operated out of the design point (i.e. best efficiency point). In this cases large coherent vortex structure (vortex rope) appears and consequently high pressure pulsations are propagated to the whole machine system leading to possible restriction of turbine operation. This flow features are consequence of flow instability called vortex breakdown in case of Francis turbine operated at part load (flow rate lower than optimal one). The present study is carried out using simplified device of swirl generator in order to access similar flow conditions as can be found in real hydraulic turbines. Both the dynamic and dissipation effect of spiral vortex breakdown are investigated. The first part of thesis deals with spiral form of vortex breakdown. The experimentally measured velocity profiles (LDA) and wall static pressures are correlated with numerical simulations carried out using open-source CFD package OpenFOAM 2.2.2. The high speed camera recording of cavitating vortex core is used to obtain image ensemble for further post-processing. The dissipation effect of spiral vortex structure is in detail discussed based on computed flow fields. The second part of thesis is dedicated to the application of POD decomposition to the study of spatio-temporal features of spiral vortex dynamics. Firstly the POD is applied to the both the experimentally obtained image ensemble of cavitating vortex and numerically computed static pressure fields. Secondly the comprehensive analysis of spiral vortex mitigation effect by the axial water jet is analyzed. The collaborative study employing the swirl generator apparatus designed by the researchers from Politehnica University of Timisoara in Romania is performed and changes in spatio-temporal vortex dynamic are studied. In this study the numerical data (in a form of three-dimensional pressure and velocity fields) are obtained using commercial CFD software ANSYS Fluent R14.

This master's thesis is about the most known methods for face recognition. There are described their advantages and disadvantages. This work is specialized at holistic methods for face recognition, which are working with 2D pictures of people. I implemented the automatic system for face recognition according to digital picture of face. There was, in this system, implemented these methods: KNN (K nearest neighbour), PCA (Principal component analysis) and LDP (Linear doscriminant projection). There was done some tests to compare implemented methods. The tests was done on the pictures from dataset FERET. In the conclusion of this text are considered implemented approaches and is marked the best method for face recognition from implemented.

In der vorliegenden Arbeit wurden 3D-Strömungszustände und die Bewegung an der Badoberfläche in Abhängigkeit von den Betriebsparametern mit der LDA-Methode im 1:2 Modell einer Stranggießkokille mit Fr-Zahl als Ähnlichkeitskriterium untersucht. Auf Basis der Messdaten wurde der obere Kokillenbereich in 7 Teilräumen stromabwärts unterteilt. Der Verlauf der Freistrahlausbreitung wurde durch eine Exponentialfunktion beschrieben. Nahe am Tauchrohraustritt wurde das Medium sowohl in den austretenden Strahl als auch in das Tauchrohr hinein eingesaugt. Die Frequenz und die Amplitude der Oberflächenschwankungen wurden nach der Leitfähigkeitsmethode gemessen. Dort dominierten drei Frequenzbereiche. In der Strömungsgeschwindigkeit beim Austritt des turbulenten Freistrahles wurden auch entsprechende nieder- und hochfrequente Anteile gemessen. Die Badoberflächenschwankungen wurden an vier Positionen gleichzeitig bestimmt. Mit steigender Fr-Zahl nahm der Mittelwert der Badoberflächenschwankung zu. Zwischen den Fr-Zahlen und den normierten Amplituden der Badoberflächenschwankungen (Mittelwert der Amplitude / hydraulischer Durchmesser der Tauchrohraustrittsfläche), konnte ein linearer Zusammenhang festgestellt werden. Die numerischen Ergebnisse (Fluent), die mit unterschiedlich definierten Randbedingungen des Strahlaustrittes berechnet wurden, ergaben eine gute Übereinstimmung mit den experimentellen Ergebnissen, wenn die Randbedingungen am Tauchrohraustritt auf experimentellen Messdaten basierten. Liegen diese Messdaten nicht vor, dann kann das Ergebnis einer numerischen Untersuchung verbessert werden, indem die stromaufwärts liegenden Systemteile in die Rechensimulation einbezogen werden.

Fortnox takes in many errands via their support chat. Some of the questions can be hard to interpret, making it difficult to know where to delegate the question further. It would be beneficial if the process was automated to answer the questions instead of need to put in time to analyze the questions to be able to delegate them. So, the main task is to find an unsupervised model that can take questions and put them into topics. A literature review over NLP and clustering was needed to find the most suitable models and techniques for the problem. Then implementing the models and techniques and evaluating them using support chat questions received by Fortnox. The unsupervised models tested in this thesis were LDA and K-means. The resulting models after training are analyzed, and some of the clusters are given a label. The authors of the thesis give clusters a label after analyzing them by looking at the most relevant words for the cluster. Three different sets of labels are analyzed and tested. The models are evaluated using five different score metrics: Silhouette, AdjustedRand Index, Recall, Precision, and F1 score. K-means scores the best when looking at the score metrics and have an F1 score of 0.417. But can not handle very small documents. LDA does not perform very well and got i F1 score of 0.137 and is not able to categorize documents together.

The web has granted everyone the opportunity to freely share large amounts of data. Individuals, corporations, and communities have made the web an important tool in their arsenal. These entities are spreading information online, but not all of it is constructive. Some spread misinformation to protect themselves or to attack other entities or ideas on the web. Checking the integrity of all the information online is a complex problem and an ethical solution would be equally complex. Multiple latent factors decide how a topic spreads and finding these factors is non-trivial. In this thesis, the patterns of different topics are compared with each other and the generalized patterns of fake, true, and mixed news, using Latent Dirichlet Allocation (LDA) topic models. We look at how the dissemination of topics can be compared through different metrics, and how these can be calculated through networks related to the data.  The analyzed data was collected using the Twitter API and news article scrapers. From this data, custom corpora were made through lemmatization and filtering unnecessary words and characters. The LDA models were made using these corpora, making it possible to extract the latent topics of the articles. By plotting the articles according to their most dominant topic, graphs for the popularity, size, and other distribution statistics could easily be drawn. From these graphs, the topics could be compared to each other and be categorized as fake, true, or mixed news by looking at their patterns and novelty. However, this brought up the question if it would be ethical to generalize topics in this way. Suppressing or censuring an article because it contains a lot of novel information might hide constructive novelties and violate freedom of speech. Finally, this thesis presents the means for further work in the future, which could involve collecting one large, continuous dataset for a fair and accurate comparison between topics.