Vi har i denna studie ämnat kartlägga hur begreppet personligt varumärke används för att återuppbygga och stärka ett personligt varumärke på Twitter av mediepersonligheter med ett comebacksyfte. Studiens teoretiska ramverk har grundats i varumärkesteorier, Erving Goffmans dramaturgiska vardagslivsperspektiv samt forskning inom begreppet personligt varumärke. Utifrån en netnografisk metod har vi samlat in och analyserat inlägg från utvalda mediepersonligheters Twitterkonto. Det empiriska materialet har bestått av en kvantitativ mängd data och därmed presenterats på ett kvantitativt sätt. Den netnografiska metoden har resulterat i en analysmodell som i sin tur har hjälpt oss analysera våra resultat utifrån olika kategorier. Analysen av våra resultat har däremot utgått ifrån en kvalitativ ansats för att förstå hur mediepersonligheterna väljer att kommunicera i sina tweets. Slutsatser Vår studie har visat på att det finns ett samband mellan det som tas upp inom begreppet personligt varumärke och mediepersonligheternas kommunikation på Twitter. Vi har utifrån vår studie dragit slutsatsen att fler följare på Twitter inte är synonymt med mest väletablerat personligt varumärke och professionellt rykte. Vidare har vi kunnat konstatera att även om en mediepersonlighets kommunikation följer principerna för att bygga ett starkt personligt varumärke men ändå inte har ett starkt varumärke betyder inte det att det ej kan stärkas med tiden. Vi har också kunnat dra slutsatsen att den personliga delen av kommunikationen, den som skapar en känsla av autenticitet och ärlighet, bör användas oftare än den säljande, ju större behovet är av att återskapa förtroende för sin publik eller stärka sitt rykte.

Trump’s new year tweet about Pakistan’s role in the fight against terrorism ignited a controversy and a war of words between Pakistani officials and the US. This thesis studies newspaper coverage of both countries on this particular issue. Dawn and The News International are chosen from Pakistan and The New York Times and The Washington Post are selected from US media. The aim of this research is to analyse and compare the media discourse. Critical discourse analysis (CDA) was adopted as the most suitable choice. A sum of eight articles, comprising two from each newspaper, is analysed intensively. Outcomes of the analysis are discussed in relation to ‘framing’ and ‘agenda setting’ theories. Five key elements from the theories are listed including: language (vocabulary), conflict presentation, sources, related issues and emphasis. Results reveal that Pakistani newspapers use very strong, rather harsh vocabulary while reporting response to Donald Trump’s tweet. The US newspapers adopted literary phrases and less harsh tone to report this issue. Conflict presentation was focused on Trump’s tweet as a central idea in all the newspapers. Pakistani newspapers focused on the coverage of reaction from military and government officials of the country. On the other hand, US newspapers included sources from both countries to have an objective view. But they have included some controversial issues, which do not have a direct link to this debate starting with Trump’s tweet. Keywords: Twitter and news media, Trump’s tweet about Pakistan, Afghan war, CDA, Framing.

Master of Science / Department of Computer Science / Doina Caragea / Huge amounts of data that are generated on social media during emergency situations are regarded as troves of critical information. The use of supervised machine learning techniques in the early stages of a disaster is challenged by the lack of labeled data for that particular disaster. Furthermore, supervised models trained on labeled data from a prior disaster may not produce accurate results. To address these challenges, domain adaptation approaches, which learn models for predicting the target, by using unlabeled data from the target disaster in addition to labeled data from prior source disasters, can be used. However, the resulting models can still be affected by the variance between the target domain and the source domain. In this context, we propose to use a hybrid feature-instance adaptation approach based on matrix factorization and the k-nearest neighbors algorithm, respectively. The proposed hybrid adaptation approach is used to select a subset of the source disaster data that is representative of the target disaster. The selected subset is subsequently used to learn accurate supervised or domain adaptation Naïve Bayes classifiers for the target disaster. In other words, this study focuses on transforming the existing source data to bring it closer to the target data, thus overcoming the domain variance which may prevent effective transfer of information from source to target. A combination of selective and transformative methods are used on instances and features, respectively. We show experimentally that the proposed approaches are effective in transferring information from source to target. Furthermore, we provide insights with respect to what types and combinations of selections/transformations result in more accurate models for the target.

Master of Science / Department of Computer Science / Doina Caragea / Twitter is one of the most active social media sites today. Almost everyone is using it, as it is a medium by which people stay in touch and inform others about events in their lives. Among many other types of events, people tweet about disaster events. Both man made and natural disasters, unfortunately, occur all the time. When these tragedies transpire, people tend to cope in their own ways. One of the most popular ways people convey their feelings towards disaster events is by offering or asking for support, providing valuable information about the disaster, and voicing their disapproval towards those who may be the cause. However, not all of the tweets posted during a disaster are guaranteed to be useful or informative to authorities nor to the general public. As the number of tweets that are posted during a disaster can reach the hundred thousands range, it is necessary to automatically distinguish tweets that provide useful information from those that don't. Manual annotation cannot scale up to the large number of tweets, as it takes significant time and effort, which makes it unsuitable for real-time disaster tweet annotation. Alternatively, supervised machine learning has been traditionally used to learn classifiers that can quickly annotate new unseen tweets. But supervised machine learning algorithms make use of labeled training data from the disaster of interest, which is presumably not available for a current target disaster. However, it is reasonable to assume that some amount of labeled data is available for a prior source disaster. Therefore, domain adaptation algorithms that make use of labeled data from a source disaster to learn classifiers for the target disaster provide a promising direction in the area of tweet classification for disaster management. In prior work, domain adaptation algorithms have been trained based on tweets represented as bag-of-words. In this research, I studied the effect of Part of Speech (POS) tag unigrams and bigrams on the performance of the domain adaptation classifiers. Specifically, I used POS tag unigram and bigram features in conjunction with a Naive Bayes Domain Adaptation algorithm to learn classifiers from source labeled data together with target unlabeled data, and subsequently used the resulting classifiers to classify target disaster tweets. The main research question addressed through this work was if the POS tags can help improve the performance of the classifiers learned from tweet bag-of-words representations only. Experimental results have shown that the POS tags can improve the performance of the classifiers learned from words only, but not always. Furthermore, the results of the experiments show that POS tag bigrams contain more information as compared to POS tag unigrams, as the classifiers learned from bigrams have better performance than those learned from unigrams.

Music ensembles have made a concerted attempt to reach out through social media platforms to the communities surrounding their concert venues in order to attract young adults to replace aging audiences. By observing opera and symphony orchestra audience members’ social media engagement through Twitter and Facebook Live, this dissertation endeavors to better understand how technology has changed the culture of classical music concert attendance. The music organizations utilizing social media considered for this study include the Los Angeles Opera, San Francisco Opera, and Boston Lyric Opera for Tweet Seats, and the Detroit Symphony Orchestra on Facebook Live. Consideration of live-tweets, Facebook Live concerts and comments, and personal interviews with social media users and music ensemble personnel provides insight to the changing experience of concert attendance. Interviews with online users who are actively participating in Tweet Seats on Twitter and chatrooms on Facebook Live during live-streamed concerts reveal that integrating social media during live performances enhances their sense of community, and their musical and social experiences. Participants indicate that prior classical music experience affects their motivation to participate and engage with other users. For many interviewees, affordability and VIP perks were initial incentives for their online involvement, but the overall experience for these users is complex. Interacting online allowed classical music fans to connect and/or reconnect to the ensembles and their music, and to an existing wired community, while negotiating with changes to the long-standing conventions of classical music culture. These alternative concert-going experiences made possible by social media reconstruct liveness within a digital world, cultivate classical music fandom, and enrich the live listening experience through collective engagement.

Opinion Mining is becoming the fastest growing area to extract useful and insightful information to support decision making. In the age of social media, user’s opinions and discussions have become a highly valuable source to look for users preferences, likes, and dislikes. The industry of wildlife parks (or zoos) is a competitive domain that requires careful analysis of visitor’s opinions to understand and cater for their preferences when it comes to wildlife. In this thesis, an opinion mining approach was proposed and applied on textual posts on the social media platform, Twitter, to extract the popularity, polarity (sentiment), and emotions toward birds and bird types such as owls, sparrows, etc. Then, the thesis provides recommendations based on popularity of birds and bird types and a ranked list of the most desired birds based on consumer emotions toward them. The findings of this thesis can help wildlife parks in the decision-making process on the types of birds to acquire.

Durant les 20 dernières années, la recherche en visualisation d’informations (InfoVis) a permis l’émergence de nouvelles techniques et méthodes qui permettent d’assister l’analyse de données intensives pour la science, l’industrie, et les gouvernements. Cependant, la plupart de ces travaux de recherches furent orientés sur des données statiques pour des utilisateurs experts.Dernièrement, des évolutions technologique et sociétales ont eu pour effet de rendre les données de plus en plus dynamiques et accessibles pour une population plus diverse. Par exemple des flux de données tels que les emails, les mises à jours de statuts sur les réseaux sociaux, les flux RSS, les systèmes de
gestion de versions, et bien d’autres. Ces nouveaux types de données sont utilisés par une population qui n’est pas forcément entraînée ou éduquée à utiliser des visualisations de données. La plupart de ces personnes sont des utilisateurs occasionnels, d’autres utilisent très souvent ces données dans leurs travaux. Dans les deux cas, il est probable que ces personnes n’aient pas reçu de formation formelle en visualisation de données.Ces changements technologiques et sociétaux ont généré une multitude de nouveaux défis, car la plupart des techniques de visualisations sont conçues pour des experts et des bases de données statiques. Peu d’études ont été conduites pour explorer ces défis. Dans ce rapport de thèse, j’adresse la question suivante : « Peut-­on permettre à des utilisateurs non­-experts de créer leur propre visualisation et de contribuer à l’analyse de flux de données ? »La première étape pour répondre à cette question est d’évaluer si des personnes non formées à la visualisation d’informations ou aux « data sciences » peuvent effectuer des tâches d’analyse de données dynamiques utiles, en utilisant un système de visualisation adapté pour supporter cette tâche. Dans la première partie de cette dissertation, je présente différents scénarios et systèmes, qui permettent à des utilisateurs non­-experts (de 20 à 300 ou 2000 à 700 000 personnes) d’utiliser la visualisation d’informations pour analyser des données dynamiques.Un autre problème important est le manque de principes génériques de design pour l’encodage visuel de visualisations d’informations dynamiques. Dans cette dissertation, je conçois, définis, et explore un espace de design pour représenter des donnés dynamiques pour des utilisateurs non­-experts. Cette espace de design est structuré par des jetons graphiques représentant des éléments de données qui permettent de construire dans le temps différentes visualisations, tant classiques que nouvelles.Dans cette thèse, je propose un nouveau paradigme de conception (design) pour faciliter la réalisation de visualisation d’informations par les utilisateurs non­-experts. Ce paradigme est inspiré par des théories établies en psychologie du développement, tout autant que par des pratiques passées et présentes de création de visualisation à partir d’objets tangibles. Je décris tout d’abord les composants et processus de bases qui structurent ce paradigme. Ensuite, j’utiliserai cette description pour étudier *si et comment* des utilisateur non­-experts sont capables de créer, discuter, et mettre à jour leurs propres visualisations. Cette étude nous permettra de réviser notre modèle précédent et de fournir une première exploration des phénomènes relatifs à la création d’encodages visuels par des utilisateurs non­-experts sans logiciel. En résumé, cette thèse contribue à la compréhension des visualisations dynamiques pour des utilisateurs non­-experts. / During the past two decades, information visualisation (InfoVis) research has created new techniques and methods to support data- intensive analyses in science, industry and government. These have enabled a wide range of analyses tasks to be executed, with tasks varying in terms of the type and volume of data involved. However, the majority of this research has focused on static datasets, and the analysis and visualisation tasks tend to be carried out by trained expert users. In more recent years, social changes and technological advances have meant that data have become more and more dynamic, and are consumed by a wider audience. Examples of such dynamic data streams include e-mails, status updates, RSS 1 feeds, versioning systems, social networks and others. These new types of data are used by populations that are not specifically trained in information visualization. Some of these people might consist of casual users, while others might consist of people deeply involved with the data, but in both cases, they would not have received formal training in information visualization. For simplicity, throughout this dissertation, I refer to the people (casual users, novices, data experts) who have not been trained in information visualisation as non-experts.These social and technological changes have given rise to multiple challenges because most existing visualisation models and techniques are intended for experts, and assume static datasets. Few studies have been conducted that explore these challenges. In this dissertation, with my collaborators, I address the question: Can we empower non-experts in their use of visualisation by enabling them to contribute to data stream analysis as well as to create their own visualizations?The first step to answering this question is to determine whether people who are not trained in information visualisation and the data sciences can conduct useful dynamic analysis tasks using a visualisation system that is adapted to support their tasks. In the first part of this dissertation I focus on several scenarios and systems where different sized crowds of InfoVis non-experts users (20 to 300 and 2 000 to 700 000 people) use dynamic information visualisation to analyse dynamic data.Another important issue is the lack of generic design principles for the visual encoding of dynamic visualization. In this dissertation I design, define and explore a design space to represent dynamic data for non-experts. This design space is structured by visual tokens representing data items that provide the constructive material for the assembly over time of different visualizations, from classic represen- tations to new ones. To date, research on visual encoding has been focused on static datasets for specific tasks, leaving generic dynamic approaches unexplored and unexploited.In this thesis, I propose construction as a design paradigm for non-experts to author simple and dynamic visualizations. This paradigm is inspired by well-established developmental psychological theory as well as past and existing practices of visualisation authoring with tangible elements. I describe the simple conceptual components and processes underlying this paradigm, making it easier for the human computer interaction community to study and support this process for a wide range of visualizations. Finally, I use this paradigm and tangible tokens to study if and how non-experts are able to create, discuss and update their own visualizations. This study allows us to refine our previous model and provide a first exploration into how non-experts perform a visual mapping without software. In summary, this thesis contributes to the understanding of dynamic visualisation for non-expert users.

Social media platforms such as Facebook and Twitter have enormous amounts of data that can be extracted and analyzed for various purposes. Stock market prediction is one of them. Previous research has shown that there is a correlation between Twitter sentiment – the proportion of positive, negative and neutral tweets – and the changes in companies’ stock prices. The present study investigates if categorizing tweets into a bigger number of categories – anger, disgust, joy, surprise, none - results in stronger correlations being found. In total, 5985 tweets in English about American Airlines, American Express, AstraZeneca and ExxonMobil were extracted and analyzed with the help of sentiment and emotion classifiers trained. Tweet sentiment showed stronger correlations with stock returns than emotion did, although the type of correlation found differed between the companies considered. It is suggested that dividing tweets into fewer categories results in semantically more distinct labels that are easier to distinguish between and that therefore show stronger correlations. Furthermore, the results indicate that the pairs of values showing the strongest correlations depend on the characteristics of each individual company.

How do people discuss mental health on social media? Can we train a computer program to recognize differences between discussions of depression and other topics? Can an algorithm predict that someone is depressed from their tweets alone? In this project, we collect tweets referencing “depression” and “depressed” over a seven year period, and train word embeddings to characterize linguistic structures within the corpus. We find that neural word embeddings capture the contextual differences between “depressed” and “healthy” language. We also looked at how context around words may have changed over time to get deeper understanding of contextual shifts in the word usage. Finally, we trained a deep learning network on a much smaller collection of tweets authored by individuals formally diagnosed with depression. The best performing model for the prediction task is Convolutional LSTM (CNN-LSTM) model with a F-score of 69% on test data. The results suggest social media could serve as a valuable screening tool for mental health.

A análise de sentimentos é um campo de estudo com recente popularização devido ao crescimento da Internet e do conteúdo que é gerado por seus usuários, principalmente nas redes sociais, nas quais as pessoas publicam suas opiniões em uma linguagem coloquial e em muitos casos utilizando de artifícios gráficos para tornar ainda mais sucintos seus diálogos. Esse cenário é observado no Twitter, uma ferramenta de comunicação que pode facilmente ser usada como fonte de informação para várias ferramentas automáticas de inferência de sentimentos. Esforços de pesquisas têm sido direcionados para tratar o problema de análise de sentimentos em redes sociais sob o ponto de vista de um problema de classificação, com pouco consenso sobre qual é o classificador com melhor poder preditivo, bem como qual é a configuração fornecida pela engenharia de atributos que melhor representa os textos. Outro problema é que em um cenário supervisionado, para a etapa de treinamento do modelo de classificação, é imprescindível se dispor de exemplos rotulados, uma tarefa árdua e que demanda esforço humano em grande parte das aplicações. Esta tese tem por objetivo investigar o uso de agregadores de classificadores (classifier ensembles), explorando a diversidade e a potencialidade de várias abordagens supervisionadas quando estas atuam em conjunto, além de um estudo detalhado da fase que antecede a escolha do classificador, a qual é conhecida como engenharia de atributos. Além destes aspectos, um estudo mostrando que o aprendizado não supervisionado pode fornecer restrições complementares úteis para melhorar a capacidade de generalização de classificadores de sentimento é realizado, fornecendo evidências de que ganhos já observados em outras áreas do conhecimento também podem ser obtidos no domínio em questão. A partir dos promissores resultados experimentais obtidos no cenário de aprendizado supervisionado, alavancados pelo uso de técnicas não supervisionadas, um algoritmo existente, denominado de C3E (Consensus between Classification and Clustering Ensembles) foi adaptado e estendido para o cenário semissupervisionado. Este algoritmo refina a classificação de sentimentos a partir de informações adicionais providas pelo agrupamento em um procedimento de autotreinamento (self-training). Tal abordagem apresenta resultados promissores e competitivos com abordagens que representam o estado da arte em outros domínios. / Sentiment analysis is a field of study that shows recent popularization due to the growth of Internet and the content that is generated by its users. More recently, social networks have emerged, where people post their opinions in colloquial and compact language. This is what happens in Twitter, a communication tool that can easily be used as a source of information for various automatic tools of sentiment inference. Research efforts have been directed to deal with the problem of sentiment analysis in social networks from the point of view of a classification problem, where there is no consensus about what is the best classifier, and what is the best configuration provided by the feature engineering process. Another problem is that in a supervised setting, for the training stage of the classification model, we need labeled examples, which are hard to get in the most of applications. The objective of this thesis is to investigate the use of classifier ensembles, exploring the diversity and the potential of various supervised approaches when these work together, as well as to provide a study about the phase that precedes the choice of the classifier, which is known as feature engineering. In addition to these aspects, a study showing that unsupervised learning techniques can provide useful and additional constraints to improve the ability of generalization of the classifiers is also carried out. Based on the promising results got in supervised learning settings, an existing algorithm called C3E (Consensus between Classification and Clustering Ensembles) was adapted and extended for the semi-supervised setting. This algorithm refines the sentiment classification from additional information provided by clusters of data, in a self-training procedure. This approach shows promising results when compared with state of the art algorithms.