Phenotypic data is of crucial importance for plant breeding in estimating a plant's biomass. Traits such as leaf area and plant height are known to be correlated with biomass. Image analysis and computer vision methods can automate data analysis for high-throughput phenotyping. Many methods have been proposed for plant phenotyping in controlled environments such as greenhouses. In this thesis, we present multiple methods to estimate traits of the plant crop sorghum from images acquired from UAV and field-based sensors. We describe machine learning techniques to extract the plots of a crop field, a method for leaf counting from low-resolution images, and a statistical model that uses prior information about the field structure to estimate the center of each plant. We also develop a new loss function to train Convolutional Neural Networks (CNNs) to count and locate objects of any type and use it to estimate plant centers. Our methods are evaluated with ground truth of sorghum fields and publicly available datasets and are shown to outperform the state of the art in generic object detection and domain-specific tasks. <br><br>This thesis also examines the use of crowdsourcing information in video analytics. The large number of cameras deployed for public safety surveillance systems requires intelligent processing capable of automatically analyzing video in real time. We incorporate crowdsourcing in an online basis to improve a crowdflow estimation method. We present various approaches to characterize this uncertainty and to aggregate crowdsourcing results. Our techniques are evaluated using publicly available datasets.<br>

Dialectal Arabic speech research suffers from the lack of labelled resources and standardised orthography. There are three main challenges in dialectal Arabic speech recognition: (i) finding labelled dialectal Arabic speech data, (ii) training robust dialectal speech recognition models from limited labelled data and (iii) evaluating speech recognition for dialects with no orthographic rules. This thesis is concerned with the following three contributions: Arabic Dialect Identification: We are mainly dealing with Arabic speech without prior knowledge of the spoken dialect. Arabic dialects could be sufficiently diverse to the extent that one can argue that they are different languages rather than dialects of the same language. We have two contributions: First, we use crowdsourcing to annotate a multi-dialectal speech corpus collected from Al Jazeera TV channel. We obtained utterance level dialect labels for 57 hours of high-quality consisting of four major varieties of dialectal Arabic (DA), comprised of Egyptian, Levantine, Gulf or Arabic peninsula, North African or Moroccan from almost 1,000 hours. Second, we build an Arabic dialect identification (ADI) system. We explored two main groups of features, namely acoustic features and linguistic features. For the linguistic features, we look at a wide range of features, addressing words, characters and phonemes. With respect to acoustic features, we look at raw features such as mel-frequency cepstral coefficients combined with shifted delta cepstra (MFCC-SDC), bottleneck features and the i-vector as a latent variable. We studied both generative and discriminative classifiers, in addition to deep learning approaches, namely deep neural network (DNN) and convolutional neural network (CNN). In our work, we propose Arabic as a five class dialect challenge comprising of the previously mentioned four dialects as well as modern standard Arabic. Arabic Speech Recognition: We introduce our effort in building Arabic automatic speech recognition (ASR) and we create an open research community to advance it. This section has two main goals: First, creating a framework for Arabic ASR that is publicly available for research. We address our effort in building two multi-genre broadcast (MGB) challenges. MGB-2 focuses on broadcast news using more than 1,200 hours of speech and 130M words of text collected from the broadcast domain. MGB-3, however, focuses on dialectal multi-genre data with limited non-orthographic speech collected from YouTube, with special attention paid to transfer learning. Second, building a robust Arabic ASR system and reporting a competitive word error rate (WER) to use it as a potential benchmark to advance the state of the art in Arabic ASR. Our overall system is a combination of five acoustic models (AM): unidirectional long short term memory (LSTM), bidirectional LSTM (BLSTM), time delay neural network (TDNN), TDNN layers along with LSTM layers (TDNN-LSTM) and finally TDNN layers followed by BLSTM layers (TDNN-BLSTM). The AM is trained using purely sequence trained neural networks lattice-free maximum mutual information (LFMMI). The generated lattices are rescored using a four-gram language model (LM) and a recurrent neural network with maximum entropy (RNNME) LM. Our official WER is 13%, which has the lowest WER reported on this task. Evaluation: The third part of the thesis addresses our effort in evaluating dialectal speech with no orthographic rules. Our methods learn from multiple transcribers and align the speech hypothesis to overcome the non-orthographic aspects. Our multi-reference WER (MR-WER) approach is similar to the BLEU score used in machine translation (MT). We have also automated this process by learning different spelling variants from Twitter data. We mine automatically from a huge collection of tweets in an unsupervised fashion to build more than 11M n-to-m lexical pairs, and we propose a new evaluation metric: dialectal WER (WERd). Finally, we tried to estimate the word error rate (e-WER) with no reference transcription using decoding and language features. We show that our word error rate estimation is robust for many scenarios with and without the decoding features.

Made available in DSpace on 2016-04-25T16:44:44Z (GMT). No. of bitstreams: 1 Milton Ruiz Rodrigues Alves.pdf: 634979 bytes, checksum: b0b282c22181c3d7e866957e3ff9e8fc (MD5) Previous issue date: 2015-05-14 / In the current time of information, we live in times that the expectations of thinkers as Lévy (1998) have become a reality. The profile of the consumer has changed as he connected with other consumers and begun to demand that there is a dialogue between the company and society. This new scenario required that technology was further developed and that the process of communication was made faster by means of online platforms that make the exchange of information viable and encourages it. The crowdsourcing arose with the availability of the crowd s intellectual capital that has changed the way companies interact and organize themselves as they come to realize the strategic importance of this information. In order to fully understand this new scenario, we have created three hypothesis that have served as guide for this paper, which are: H1 Crowdsourcing affects products performance; H2 negative reviews from crowd voting provide the same effects that positive reviews; and H3- Crowd voting provides information about consumer`s perception of the product being offered. Therefore, this paper aims to understand if the crowdsourcing affects product s performance of the companies that use crowdsourcing by analyzing the theoretical basis of crowdsourcing and also by reviewing a data-base with reviews made by consumers during a four year period. In addition, this work also aims at raising questions for new researches that may develop further the understanding of the effects of the crowdsourcing. Based on the review of the correlations between the variables obtained in Amazon s data base and the analysis of the feelings used in the word clouds based on the reviews offered by the consumers, this work suggests that the crowdsourcing does affect the product s performance / Vivemos num momento em que as expectativas de pensadores como Lévy (1998), se concretizaram. O consumidor mudou, passou a exigir um diálogo entre a empresa e a sociedade. Esse novo cenário cobrou da tecnologia formas de aperfeiçoar e agilizar este processo de comunicação, plataformas baseadas na internet que propiciem e incentivam a troca de informações. Nesta esfera surgiu o crowdsourcing, com a disponibilização do capital intelectual das massas, que acabou por mudar a forma de interagir e se organizar das empresas que cada vez mais percebe a importância estratégica destas informações. Para compreender este novo cenário, elaborou-se três hipóteses que serviram de guia para a dissertação, sendo estas: H1 Crowdsourcing afeta a performance de produtos; H2 Comentários negativos originados pelo crowd voting apresentam os mesmos efeitos que os comentários positivos e H3- Crowd voting fornece dados sobre o que está sendo percebido pelo mercado consumidor frente ao produto anunciado. Objetivando assim compreender a partir da fundamentação teórica do crowdsourcing e a análise de um banco de dados com dados de avaliações realizadas por usuários ao longo de quatro anos se o crowdsourcing causa impactos na performance dos produtos ofertados por empresas que o pratiquem, e desta forma abrir caminhos para novas pesquisas, onde possa-se sugerir caminhos para o aprofundamento do conhecimento teórico e prático desta arte. A partir disto, mediante o estudo das correlações entre as variáveis obtidas pelo banco de dados da Amazon e a análise de sentimentos utilizada na formulação de nuvens de palavras feitas com os comentários fornecidos pelos usuários sobre os produtos, o estudo sugere que o crowdsourcing exerce sim impactos na performance de produtos das empresas

O presente trabalho versa sobre as narrativas e justificativas relativas ao processo de criação de um mercado de financiamentos coletivos no Brasil a partir de diversos interlocutores: criadores de projetos, criadores de websites e a partir da literatura acadêmica mais comum sobre o tema. Busca-se retratar e discutir a criação deste mercado a partir do estabelecimento de um passado e de uma pedagogia, típicas dos circuitos de consumo relativos ao fenômeno dos crowdfunding, de modo a circunscrever o que meus interlocutores chamam de “cultura do financiamento coletivo”. Este mercado constitui-se em um duplo processo de afastamento e aproximação entre plataformas como forma de delimitar moralmente o que “deve”, ou não, ser este fenômeno. Tal processo revela a emergência de controvérsias, agenciamentos e, principalmente, de sujeitos como “empreendedores” que buscam “fazer a diferença” a partir de projetos e plataformas de financiamento coletivo. No limite, os financiamentos coletivos alinham-se à elaboração de justificativas críticas sobre o papel do Estado e de empresas na constituição dos mercados, ainda que reiterem práticas comuns no campo empresarial, como a manipulação publicitária e a tentativa de criação de vínculos – ou attachments – com o público alvo. Assim, doação e pré-compra confundem-se, bem como as motivações para criação de projetos de financiamento coletivo. Teoricamente, este trabalho sustenta-se nas digressões da Sociologia Pragmática de Luc Boltanski e da compreensão sobre mercados oriunda da Sociologia e Antropologia, principalmente inspirada nos escritos de Michel Callon e Jens Beckert. Busca-se contribuir para o campo da Antropologia a partir da discussão sobre o processo de criação de mercados a partir dos sujeitos que dele fazem parte e de suas narrativas. / This dissertation discusses the narratives and justifications related to the process of creating a market for crowdfunding in Brazil based on the people who create projects and websites, mainly, as well as the most common literature in the field. We intend to portray and discuss the establishment of this market through its past and pedagogy, typical of circuits of commerce in Crowdfunding whose actors intend to circumscribe what they call as a “Crowdfunding culture”. This market is comprised of a double-movement of approximation and distance among platforms as a way to morally determine what is supposed to be a Crowdfunding and what is not supposed to be. This process reveals the emergence of controversies, agencies and, mainly, entrepreneurial subjects that aim at “making the difference” through Crowdfunding platforms and projects. Ultimately, the Crowdfunding is related to the creation of critical justification on the role of the Government and Private sector in the constitution of markets, even if Crowdfunding itself recall common corporate practices as communication strategies, for instance, which aim at creating attachments with the target audience. Therefore, donation and pre-buying mechanisms overlap, as well as the motivations to create such projects. Theoretically, this work is supported by the Pragmatic Sociology of Luc Boltanski and the comprehension of markets based on Sociology and Anthropology, mainly inspired by the point of view of Michel Callon and Jens Beckert. This dissertation intends to contribute to Anthropology by discussing the process of market creation from the point of view of the subjects that work in it and its narratives.

Made available in DSpace on 2016-04-29T14:23:21Z (GMT). No. of bitstreams: 1 Daniel Calado Nunes e Souza.pdf: 1220555 bytes, checksum: 44ff978e253d62ac6964d520cf9b50e8 (MD5) Previous issue date: 2013-06-07 / Project research and practical application to stimulate the binding of dramaturgical audiovisual creations generated with collective participation through the internet, by proposing a linkageby virtual interconnection of scenesrecorded separatelyby different participants in different geographical locations, building up scenes as modules ready for audiovisual crossed montages, multireferred andmultilinear, forming a dynamic network. The process of collective creation was suggested for all stages of creation, since the invention of drama, action and coming through the montagemade by the participants themselves, seeking presential action, including the creationbybodilymanifestation, and virtual integration, mediated and published overthe Internet. The intent is that participants invent narratives or audiovisual associations bynegotiating ofdramaturgical elements present in their scenes. Was taken as the reference multireferenciality generated by the encyclopedic natureof some productions, like the Matrix movies, StarWars and the series of Harry Potter, the allegories ofBrazilian moviesof the era of Marginal Cinema and theater productions and films open to collective creation. Emphasis was placed on the phenomenon of media convergence generated by a common binary system, facilitating the invention of matrices preparedfor sharedartistic production, and was broughttheillustration of hypermedia like a network of lexias connected by links to propose ways of audiovisual mounting that take the links as audiovisual transitions and sopromotesthe networked montage. This proposal enjoys the concept of agency widely described by Janet Murray, referring to the ability and willingness of the participants to interfere in works. It was consideredthat the Internet still has a large field of evolution for originalexperiences of collective creation. We seek analogies between the new possibilities of action brought by the internet and the participatory theater developed by theater director Augusto Boal, and in the field of montage weresorted toanalysis of Xavier on the history of film editing, linking it to the research of Lev Manovich and Vicente Gosciola on the production and montagefor hypermedia. These questions were elucidatedfront ofthe principle of collective intelligence as advocatedby Pierre Lévy. As a result, we designed a prototype exemplifyingthe project. In the final analysis, we proposed some changes in video sharing platforms to fostera culture of collective montage on the internet, while we encourage the promotion of leisure made by its own participants and ways of collaborative artistic production / Projeto de pesquisa e aplicação prática para estimular a vinculação de criações dramatúrgicas audiovisuais geradascom participação coletiva atravésda internet, propondo a vinculação pela interligação virtual deencenações gravadas isoladamente, por diferentes participantes em diferentes locais geográficos, construindo-secenas como módulos preparados para montagens audiovisuaiscruzadas, multirreferenciadas e multilineares, compondo uma rede dinâmica. O processo de criação coletiva foi sugerido para todas as etapas de criação, desde a invenção dramatúrgica, passando pela atuação e chegando à montagem feita pelos próprios participantes, buscando a ação presencial, inclusive a criação pela manifestação corporal, e a integração virtual, mediada e divulgada pela internet. O intuito é que os participantes inventem narrativas ouassociações audiovisuais pela negociação de elementos dramatúrgicos e audiovisuais presentes em suas cenas. Tomou-se como referência a multirreferencialidade gerada pelo caráter enciclopédico de algumas obras, como os filmes Matrix, StarWars e as séries de Harry Potter, a recorrência a alegorias em filmes brasileiros da era do Cinema Marginal, e produções teatrais e audiovisuais abertas à criação coletiva. Ressaltou-se o fenômeno da convergência de mídias gerado por um sistema binário comum, facilitando a invenção de matrizes para a produção artística compartilhada, e a ilustração da hipermídia como uma rede de lexias ligadas por links para propor formas de montagem audiovisual que tomam os links como formas de transição audiovisual, promovendo a montagem em rede. Tal proposta usufrui amplamente do conceito de agência descrito por Janet Murray, referindo-se à possibilidade e à vontade dos participantes de interferirem numa obra. Considerou-se ainda que a internet apresenta um grande campo de evolução para experiências de criação coletiva originais. Buscamos analogias entre as novas possibilidades de ação trazidas pela internet e o teatro participativo desenvolvido pelo diretor teatral Augusto Boal, e no campo da montagem recorremos às análises de Ismail Xavier sobre a história da montagem cinematográfica, relacionando-a às pesquisas de Lev Manovich e Vicente Gosciola sobre a produção e montagem para a hipermídia. Essas questões foram elucidadas diante do princípio da inteligência coletiva como defendido por Pierre Lévy. Como resultado, foi elaborado um protótipo exemplificando o projeto. Numa análise final, propusemos algumas alterações em plataformas de compartilhamento de vídeos para promover uma cultura de montagem coletiva na internet, ao mesmo tempo que incentivamos a promoção coletiva do próprio lazer e da produção artística colaborativa

Submitted by PPG Ci?ncia da Computa??o (ppgcc@pucrs.br) on 2018-10-24T18:17:00Z No. of bitstreams: 1 ALEXANDRE LAZARETTI ZANATTA.DIS.pdf: 3682553 bytes, checksum: f4e0c608791ce6787d609d8099456e04 (MD5) / Approved for entry into archive by Sheila Dias (sheila.dias@pucrs.br) on 2018-10-26T11:47:02Z (GMT) No. of bitstreams: 1 ALEXANDRE LAZARETTI ZANATTA.DIS.pdf: 3682553 bytes, checksum: f4e0c608791ce6787d609d8099456e04 (MD5) / Made available in DSpace on 2018-10-26T11:56:21Z (GMT). No. of bitstreams: 1 ALEXANDRE LAZARETTI ZANATTA.DIS.pdf: 3682553 bytes, checksum: f4e0c608791ce6787d609d8099456e04 (MD5) Previous issue date: 2018-08-13 / Software crowdsourcing development platforms require a continuous influx of crowdworkers for their continuity. Crowdworkers should be encouraged to play an important role in the online communities by being active members, but they face difficulties when attempting to participate. For this reason, we investigated the difficulties that crowdworkers face in crowdsourcing software development platforms. We conducted empirical studies relying on multiple data sources and research methods including literature review, peer review, field study, and procedures of grounded theory. We observed that crowdworkers face many barriers ? related to competence, collaboration, and time management ? when making their contributions in software crowdsourcing development, which can result in dropouts. Based on the identified barriers, literature review and, crowdworkers suggestions, we list 13 recommendations for participants as potential solutions to overcome such barriers. The recommendations were evaluated by surveying software crowdsourcing experts. The main contributions of this dissertation are a) empirical identification of barriers faced by crowdsourcing software development crowdworkers; and b) recommendations on how to minimize the barriers. We conclude that the crowdworkers need competency and an efficient time management effort to take part collaboratively in tasks of the Competition-Based Crowdsourcing Software Development of the Topcoder platform. / O desenvolvimento de projetos por meio de software crowdsourcing depende de um fluxo cont?nuo de trabalhadores da multid?o para sua continuidade. Estes trabalhadores necessitam ser membros ativos, por?m enfrentam dificuldades quando tentam participar em projetos de software crowdsourcing. Por esta raz?o, investigou-se quais foram as dificuldades que os trabalhadores da multid?o enfrentam em plataformas de desenvolvimento de projetos via software crowdsourcing. Foram realizados v?rios estudos emp?ricos baseados em m?ltiplas fontes de dados e m?todos de pesquisa, incluindo revis?o da literatura, revis?o por pares, estudo de campo e procedimentos da teoria fundamentada. Observou-se que os trabalhadores enfrentam muitas barreiras - relacionadas ? compet?ncia, colabora??o e gerenciamento do tempo - ao fazerem suas contribui??es no desenvolvimento projetos por software crowdsourcing, o que pode resultar em desist?ncias. Com base nas barreiras identificadas, revis?o de literatura e sugest?es dos trabalhadores foram apresentadas 13 recomenda??es para que estes trabalhadores possam superar tais barreiras. Estas recomenda??es foram avaliadas por especialistas em software crowdsourcing. As principais contribui??es desta tese foram: a) identifica??o emp?rica das barreiras enfrentadas pelos desenvolvedores de software em projetos de software crowdsourcing; e b) recomenda??es para minimizar estas barreiras. Conclui-se que os trabalhadores da multid?o precisam de compet?ncia e um esfor?o eficiente de gerenciamento de tempo para participar de forma colaborativa nas tarefas do desenvolvimento de projetos em software Crowdsourcing no modelo competitivo da plataforma Topcoder.

Measurement collection is a primary step towards analyzing and optimizing performance of a telecommunication service. With an Mobile Broadband (MBB) network, the measurement process has not only to track the network's Quality of Service (QoS) features but also to asses a user's perspective about its service performance. The later requirement leads to "user-side measurements" which assist in discovery of performance issues that makes a user of a service unsatisfied and finally switch to another network. User-side measurements also serve as first-hand survey of the problem domain. In this thesis, we exhibit the potential in the measurements collected at network edge by considering two well-known approaches namely crowdsourced and distributed testbed-based measurements. Primary focus is on exploiting crowdsourced measurements while dealing with the challenges associated with it. These challenges consist of differences in sampling densities at different parts of the region, skewed and non-uniform measurement layouts, inaccuracy in sampling locations, differences in RSS readings due to device-diversity and other non-ideal measurement sampling characteristics. In presence of heterogeneous characteristics of the user-side measurements we propose how to accurately detect mobile coverage holes, to devise sample selection process so to generate a reliable radio map with reduced sample cost, and to identify cellular infrastructure at places where the information is not public. Finally, the thesis unveils potential of a distributed measurement test-bed in retrieving performance features from domains including user's context, service content and network features, and understanding impact from these features upon the MBB service at the application layer. By taking web-browsing as a case study, it further presents an objective web-browsing Quality of Experience (QoE) model.

Although small businesses borrowed $1 trillion in 2013 from traditional lenders, 35% of small business owners were unable to obtain adequate financing and subsequently sought alternative sources such as crowdfunding. Guided by the pecking order theory, the purpose of this exploratory case study was to explore how 6 small business owners in Tennessee successfully used crowdfunding to start, grow, or sustain their businesses. Data were collected from semistructured interviews and a review of crowdfunding project data on the internet platform including the project description, target goal, amount achieved, number of backers, and locations of the funders. Data were inductively analyzed, first into coded phrases, then categories, and finally emergent themes. Findings revealed that these small business owners tapped into a strong social media network of potential funders for increased funding opportunities. They also advocated that project descriptions consist of high-quality project content and videos, 9 to 11 reward levels, and valuable rewards to entice funders to contribute to the campaign. These small business owners also noted that they devoted more time than originally anticipated during the planning, execution, and fulfillment phases, and they all faced preliminary transaction, fulfillment, and shipping costs when using crowdfunding. The risks included not receiving any funding, negative customer feedback, and poor reputation. These stories have implications for positive social change by illuminating the necessary resources to establish a successful business through employment of a social change mechanism. With funding for growth, the small business owner, family, and local community will promote economic prosperity.

Physicians must participate in developing medical protocols to ensure that medical best practices are adopted for patients' social benefit. Healthcare leaders have struggled to gain sufficient physician participation in developing medical protocols. Using technology-based crowdsourcing to assimilate knowledge from physicians may help healthcare managers improve medical protocol development. Using self-determination theory, this quantitative causal-comparative design aimed to determine whether differences in intrinsic and extrinsic motivation existed among the 132 participating physicians who did or did not participate in developing medical protocols in a crowdsourcing environment. Participants were recruited by e-mail through an independent physician association. Motivation levels were measured by the Aspirations Index via an online survey. A total of 55.3% of respondents participated in developing medical protocols. Differences were anticipated in the levels of participation in developing medical protocols between intrinsically and extrinsically motivated physicians. Rank correlations were computed between the number of protocols completed and all of the motivation scores. Personal growth and community contribution were significantly correlated with the number of addressed protocols. Positive social change may occur through improving medical protocols and healthcare outcomes by informing healthcare leaders about physicians' motivation to participate in developing medical protocols. By understanding these motivators, leaders can highlight the benefits of protocol development to encourage physician participation. If participation is enhanced, protocol quality and healthcare effectiveness may be improved, benefitting patients and healthy individuals.

Crowdsourcing systems enable new opportunities for requesters with limited funds to accomplish various tasks using human computation. However, the power of human computation is abused by malicious requesters who create malicious campaigns to manipulate information in web systems such as social networking sites, online review sites, and search engines. To mitigate the impact and reach of these malicious campaigns to targeted sites, we propose and evaluate a machine learning based classification approach for detecting malicious campaigns in crowdsourcing platforms as a first line of defense, and build a malicious campaign blacklist service for targeted site providers, researchers and users. Specifically, we (i) conduct a comprehensive analysis to understand the characteristics of malicious campaigns and legitimate campaigns in crowdsourcing platforms, (ii) propose various features to distinguish between malicious campaigns and legitimate campaigns, (iii) evaluate a classification approach against baselines, and (iv) build a malicious campaign blacklist service. Our experimental results show that our proposed approaches effectively detect malicious campaigns with low false negative and false positive rates.