The mainstream quantitative models in the finance literature have been ineffective in detecting possible bankruptcies during the 2007 to 2009 financial crisis. Coinciding with the same period, various researchers suggested that sentiments in social media can predict future events. The purpose of the study was to examine the relationship between investor sentiment within the social media and the financial distress of firms Grounded on the social amplification of risk framework that shows the media as an amplified channel for risk events, the central hypothesis of the study was that investor sentiments in the social media could predict t he level of financial distress of firms. Third quarter 2014 financial data and 66,038 public postings in the social media website Twitter were collected for 5,787 publicly held firms in the United States for this study. The Spearman rank correlation was applied using Altman Z-Score for measuring financial distress levels in corporate firms and Stanford natural language processing algorithm for detecting sentiment levels in the social media. The findings from the study suggested a non-significant relationship between investor sentiments in the social media and corporate financial distress, and, hence, did not support the research hypothesis. However, the model developed in this study for analyzing investor sentiments and corporate distress in firms is both original and extensible for future research and is also accessible as a low-cost solution for financial market sentiment analysis.

Com o avanço da Internet pelo mundo, as pessoas passaram a interagir cada vez mais com a Web, principalmente após o surgimento das redes sociais, criando conteúdo que pode ser explorado de diversas formas. Esse aumento de usuários tem sido global, ou seja, pessoas de diversos países passaram a produzir textos de diversos idiomas. Esses textos compõem um rico conteúdo para Análise de Sentimentos Multilíngue. A maior parte dos trabalhos da área se foca em Mineração de Opinião, analisando o sentimento através da polaridade. Outro tipo de sentimento que tem atraído atenção é a emoção, embora não seja amplamente explorada a Análise de Sentimentos Multilíngue usando emoção. Este trabalho utiliza técnicas geralmente usadas para Mineração de Opinião e polaridade para Análise de Sentimentos Multilíngues usando emoção. O objetivo deste trabalho é comparar diferentes combinações de aprendizado de máquina supervisionado e tradução automática para criar corpora em diferentes idiomas a partir de corpora anotados já existentes. As duas formas de utilizar as traduções comparadas são: criando classificadores de emoção separados por idiomas, chamados monolíngues, e criando um classificador composto do idioma original e das traduções, chamado multilíngue. É feito ainda um experimento cruzando dois corpora, visando avaliar o uso da tradução de um corpus com os textos originais do outro. Os resultados dos experimentos mostram não apenas o sucesso de analisar emoção usando aprendizado supervisionado e tradução automática, mas que o classificador multilíngue supera os classificadores monolíngues. O experimento cruzando os corpora mostra que para algumas emoções os corpora estão alinhados, mas que para outras é preciso que haja maior similaridade nos textos. / With the growth of the Internet around the world, people began to interact more and more with the Web, especially after the emergence of social networks, creating content that can be exploited in several ways. This increase in the number of users has been global, that is, people from different countries started producing texts in several languages. These texts comprise a rich content for Multilingual Sentiment Analysis. Most of the work in the area focus in Opinion Mining, analyzing the feeling through polarity. Another type of feeling that has attracted attention is emotion, although not extensively explored in Multilingual Sentiment Analysis. This work uses techniques commonly used for Opinion Mining and polarity for Multilingual Sentiment Analysis using emotion. The objective of this study is to compare different combinations of supervised machine learning and automatic translation to create corpora in different languages from existing annotated corpora. The two ways to use the translations compared are: creating emotion classifiers separated by languages, called monolingual, and creating a composed classifier, with the original language and it’s translations, called multilingual. An experiment crossing the two corpora used is made, to evaluate the use of the translation of one corpus with the original texts of the other. The results of the experiments show not only the success of analysing emotion using supervised machine learning and automatic translation, but that the multilingual classifier exceeds the monolingual classifiers. The experiment crossing the corpora shows that to some emotions the corpora are aligned, but for others there needs to be greater similarity in the texts.

Cette thèse étudie plusieurs aspects de la dynamique du rendement des actifs. Les trois premiers chapitres ont pour objet la formation des prix sur le marché de l'art. Le premier chapitre établit que les prix peuvent s'écarter temporairement, et de manière partiellement prévisible, de la valeur fondamentale. Cet article a été publié dans Economics Letters (Volume 122, Issue 3, pp. 432-434) et a été écrit avec Christophe Spaenjers et Luc Renneboog. Le chapitre 2 étudie la vitesse de transmission de l'information dans les prix agrégés du marché de l'art. Le chapitre 3 analyse la corrélation entre prix et volume et étaye des éléments concordant avec une hypothèse de bulles. Il a été écrit avec Luc Renneboog. Le chapitre 4 s'attache à la modélisation empirique de la prédictibilité d'indices boursiers sur quinze pays industrialisés. Il propose de combiner l'information donnée par chaque pays de façon à améliorer le pouvoir prédictif. / The doctoral thesis studies several aspects of asset returns dynamics. The first three chapters focus on returns in the fine art market. The first chapter provides evidence for the existence of a slow-moving fad component in art prices that induces short-term return predictability. The article has been published in Economics Letters (Volume 122, Issue 3, pp. 432-434), and was written together with Christophe Spaenjers and Luc Renneboog. Chapter 2 investigates how fast is information incorporated into aggregate art prices. Chapter 3 studies price-volume dynamics in the art market and documents evidence of bubble patterns in prices and is written with Luc Renneboog. Chapter 4 proposes a Bayesian estimation procedure that makes efficient use of cross-sectional information, and revisits the return predictability literature.

Notre recherche a pour objectif de présenter le concept de la justice chez Rousseau dans sa complexité et ses contradictions apparentes. La justice ne se pense ni en termes proprement moraux ni en termes politiques. Il s’agit plutôt d’une articulation de la politique et de la morale, articulation qui se fait, par contre, à partir d’un sens d’abord moral qui peut jouer alors le rôle fondateur d’une théorie possible de la justice ; « sens de la justice » donc, d’abord élaboré par Rousseau, et théorisé par John Rawls qui en fait l’un des maillons clefs de sa généalogie. C’est donc seulement en comprenant les motivations qui fondent une société, c’est-à-dire en cherchant la genèse de ce sens, que l’on pourra établir ensuite les principes de justice. Dans l’Émile, Rousseau interroge le sens de la justice en affirmant, d'une manière originale, que le sentiment de la justice est historiquement antérieur au raisonnement à propos de la justice. En partant de la définition de la pitié comme sentiment naturel à l'être humain, le rendant sensible aux souffrances des autres, il met l'accent sur l'importance de ce sentiment dans la formation de l'idée de la justice. Par ailleurs, dans son contrat social, Rousseau montre l’autre aspect de la justice, la justice en tant qu'elle est amour de l'ordre ; amour du corps politique. En ce sens, la justice est une harmonie entre tous les membres du corps politique où chacun est à la fois membre du souverain et sujet. Notre volonté d'analyser la notion de la justice chez Rousseau nous conduit à l'examen d'une série de questions. Voici les plus importantes : qu’est-ce que le sens de la justice chez Rousseau ? Sa doctrine du sens moral de la justice est-elle en opposition avec sa théorie politique et sa conception de la volonté générale ? Est-il possible, malgré cette dualité de principes et de définitions, de découvrir un critère commun, sentimental ou rationnel de la justice ? / Our research aims to present the concept of justice in Rousseau in its complexity and apparent contradictions. The justice cannot be understood in strictly moral or political terms. It is rather an articulation of politics and morality, which is based on a sense of morality that can then plays the founding role of a possible theory of justice; “sense of justice” therefore, first elaborated by Rousseau and then theorized by John Rawls who makes it one of the key links of his genealogy. It is only while understanding motivations that founded society, that is to say, by looking for the origin of this “sense”, that the principles of justice can then be established. In Émile, Rousseau questions the meaning of justice by asserting, in an original way, that the feeling of justice is historically prior to reasoning about justice. Starting from the definition of pity as a natural feeling to the human being, making him sensitive to the suffering of others, he emphasizes the importance of this feeling in the formation of the idea of justice. Moreover, in his social contract, Rousseau shows the other aspect of justice, justice as love of the order; the love of the body politic. In this sense, justice is a harmony between all the members of the body politic where everyone is at the same time member of the sovereign and subject. Our desire to analyze Rousseau’s notion of justice leads us to examine a series of questions. Here are the most important : what is Rousseau's sense of justice? Is his doctrine of the moral sense of justice in opposition to his political theory and his conception of the general will ? Is it possible, despite this duality of principles and definitions, to discover a common, sensitive or rational criterion of justice?

Com o avanço da Internet pelo mundo, as pessoas passaram a interagir cada vez mais com a Web, principalmente após o surgimento das redes sociais, criando conteúdo que pode ser explorado de diversas formas. Esse aumento de usuários tem sido global, ou seja, pessoas de diversos países passaram a produzir textos de diversos idiomas. Esses textos compõem um rico conteúdo para Análise de Sentimentos Multilíngue. A maior parte dos trabalhos da área se foca em Mineração de Opinião, analisando o sentimento através da polaridade. Outro tipo de sentimento que tem atraído atenção é a emoção, embora não seja amplamente explorada a Análise de Sentimentos Multilíngue usando emoção. Este trabalho utiliza técnicas geralmente usadas para Mineração de Opinião e polaridade para Análise de Sentimentos Multilíngues usando emoção. O objetivo deste trabalho é comparar diferentes combinações de aprendizado de máquina supervisionado e tradução automática para criar corpora em diferentes idiomas a partir de corpora anotados já existentes. As duas formas de utilizar as traduções comparadas são: criando classificadores de emoção separados por idiomas, chamados monolíngues, e criando um classificador composto do idioma original e das traduções, chamado multilíngue. É feito ainda um experimento cruzando dois corpora, visando avaliar o uso da tradução de um corpus com os textos originais do outro. Os resultados dos experimentos mostram não apenas o sucesso de analisar emoção usando aprendizado supervisionado e tradução automática, mas que o classificador multilíngue supera os classificadores monolíngues. O experimento cruzando os corpora mostra que para algumas emoções os corpora estão alinhados, mas que para outras é preciso que haja maior similaridade nos textos. / With the growth of the Internet around the world, people began to interact more and more with the Web, especially after the emergence of social networks, creating content that can be exploited in several ways. This increase in the number of users has been global, that is, people from different countries started producing texts in several languages. These texts comprise a rich content for Multilingual Sentiment Analysis. Most of the work in the area focus in Opinion Mining, analyzing the feeling through polarity. Another type of feeling that has attracted attention is emotion, although not extensively explored in Multilingual Sentiment Analysis. This work uses techniques commonly used for Opinion Mining and polarity for Multilingual Sentiment Analysis using emotion. The objective of this study is to compare different combinations of supervised machine learning and automatic translation to create corpora in different languages from existing annotated corpora. The two ways to use the translations compared are: creating emotion classifiers separated by languages, called monolingual, and creating a composed classifier, with the original language and it’s translations, called multilingual. An experiment crossing the two corpora used is made, to evaluate the use of the translation of one corpus with the original texts of the other. The results of the experiments show not only the success of analysing emotion using supervised machine learning and automatic translation, but that the multilingual classifier exceeds the monolingual classifiers. The experiment crossing the corpora shows that to some emotions the corpora are aligned, but for others there needs to be greater similarity in the texts.

Les réseaux sociaux numériques ont pris une place prépondérante dans l'espace informationnel, et sont souvent utilisés pour la publicité, le suivi de réputation, la propagande et même la manipulation, que ce soit par des individus, des entreprises ou des états. Alors que la quantité d'information rend difficile son exploitation par des humains, le besoin reste entier d'analyser un réseau social numérique : il faut dégager des tendances à partir des messages postés dont notamment les opinions échangées, qualifier les comportements des utilisateurs, et identifier les structures sociales émergentes.Pour résoudre ce problème, nous proposons un système d'analyse en trois niveaux. Tout d'abord, l'analyse du message vise à en déterminer l'opinion. Ensuite, la caractérisation et l'évaluation des comptes utilisateurs est réalisée grâce à une étape de profilage comportemental et à l'étude de leur importance et de leur position dans des graphes sociaux, dans lesquels nous combinons les mesures topologiques d'importance des noeuds dans un graphe avec les statistiques d'engagement, par exemple en nombre d'abonnés. Enfin, le système procède à la détection et à l'évaluation de communautés d'utilisateurs, pour lesquelles nous introduisons des scores de cohésion thématique qui complètent les mesures topologiques classiques de qualité structurelle des communautés détectées. Nous appliquons ce système d'analyse sur deux corpus provenant de deux médias sociaux différents : le premier est constitué de messages publiés sur Twitter, représentant toutes les activités réalisées par 5 000 comptes liés entre eux sur une longue période. Le second provient d'un réseau social basé sur TOR, nommé Galaxy2. Nous évaluons la pertinence de notre système sur ces deux jeux de données, montrant la complémentarité des outils de caractérisation des comptes utilisateurs (influence, comportement, rôle) et des communautés de comptes (force d'interaction, cohésion thématique), qui enrichissent l'exploitation du graphe social par les éléments issus des contenus textuels échangés. / Online Social Networks have taken a huge place in the informational space and are often used for advertising, e-reputation, propaganda, or even manipulation, either by individuals, companies or states. The amount of information makes difficult the human exploitation, while the need for social network analysis remains unsatisfied: trends must be extracted from the posted messages, the user behaviours must be characterised, and the social structure must be identified. To tackle this problem, we propose a system providing analysis tools on three levels. First, the message analysis aims to determine the opinions they bear. Then, the characterisation and evaluation of user accounts is performed thanks to the union of a behavioural profiling method, the study of node importance and position in social graphs and engagement and influence measures. Finally the step of user community detection and evaluation is accomplished. For this last challenge, we introduce thematic cohesion scores, completing the topological, graph-based measures for group quality. This system is then applied on two corpora, extracted from two different online social media. The first is constituted of messages published on Twitter, gathering every activity performed by a set of 5,000 accounts on a long period. The second stems from a ToR-based social network, named Galaxy2, and includes every public action performed on the platform during its uptime. We evaluate the relevance of our system on these two datasets, showing the complementarity of user account characterisation tools (influence, behaviour and role), and user account communities (interaction strength, thematic cohesion), enriching the social graph exploitation with textual content elements.

Sentiment classification is the process of categorizing data into categories based on its polarity with a wide array of applications across several industries. This report examines a combination of two prominent approaches to sentiment classification using a lexicon of weighted words and machine learning respectively. These approaches are compared with the combined hybrid approach in order to give an account of their relative strengths and weaknesses. When run on a set of IMDb movie reviews the results indicate that the hybrid model performs better than the lexicon-based approach, in turn being outperformed by the learning-based approach. However, the gain in convenience brought on by eliminating the need for training data makes the hybrid model an appealing alternative to the other approaches with a slight trade-off in performance. / Att klassificera text i kategorier baserat på känslan de uttrycker är ett aktuellt område idag och kan tillämpas inom många industrier. Rapporten undersöker en kombination av de två framstående tillvägagångssätten till denna typ av klassificering baserade på ett lexikon med definerade ordvikter respektive maskininlärning. Denna hybridlösning jämförs mot de två andra tillvägagångssätten för att framlägga deras relativa styrkor och svagheter. På ett dataset med filmrecensioner från IMDb får maskininlärningsklassificeraren bäst resultat, följt av hybridlösningen och sist den lexikonbaserade lösningen. Trots det kan hybridlösningen vara att föredra i situationer där det är ogenomförbart eller oskäligt att förbereda träningsdata för maskininlärningsklassificeraren, dock med ett visst avkall på prestanda.

No description available.

The presented dissertation examines the potential of using the sentiment score extracted from textual data with historical stock index data to improve the performance of stock market prediction through the created model of the expert system. Given the large number of financial-related text documents published by both professional and amateur investors, not only on online social networks that could have an impact on real stock markets, but it is also crucial to analyze and in particular extract financial texts published by different users. investor sentiment. In this work, investor sentiment is obtained from online financial reports and contributions published on the financial social platform StockTwits. Sentiment scores are determined using a hybrid approach combining machine learning models with the teacher and neural networks, with multiple lexicons of positive and negative words used to classify sentiment polarity. The influence of sentiment score on the stock market through causality, cointegration and coherence is analyzed. The dissertation proposes a model of an expert system based on fuzzy logic methods. Fuzzy logic provides remarkable features when working with vague, inaccurate or unclear data and is able to deal with the chaotic environment of stock markets. In recent scientific studies, it has gained in popularity a higher level of fuzzy logic, which is referred to as type-2 fuzzy logic. Unlike the classic type-1 fuzzy logic, this higher type is able to integrate a certain level of uncertainty between the dual membership functions. However, this type of expert system is considerably neglected in the subject issue of stock market prediction using the extracted investor sentiment. For this reason, the dissertation examines the potential to use and the performance of type-2 fuzzy logic. Specifically, several type-2 fuzzy models are created. which are trained on historical stock index data and sentiment scores extracted from text data for the period 2018-2020. The created models are assessed to measure the prediction performance without sentiment and with the integration of investor sentiment. Subsequently, based on the created expert model, the investment strategy is determined, and its profitability is monitored. The prediction performance of fuzzy models is compared with the performance of several comparison models, including SVM, KNN, naive Bayes and others. It has been observed from experiments that fuzzy logic models are able to improve prediction by appropriate setting of membership and uncertainty functions contained in them and are able to compete with classical expert prediction models, which are standardly used in research studies. The created model should serve as a tool to support investment decisions for individual investors.

General purpose Twitter sentiment analysis is a novel field that is closely related to traditional Twitter sentiment analysis but slightly differs in some key aspects. The main difference lies in the fact that the novel approach considers the unfiltered public Twitter stream while most of the previous approaches often applied various filtering steps which are not feasible for many applications. Another goal is to yield more reliable results by only classifying a tweet as positive or negative if it distinctly consists of the respective sentiment and mark the remaining messages as uncertain. Traditional approaches are often not that strict. Within the course of this thesis it could be verified that the novel approach differs significantly from the traditional approach. Moreover, the experimental results indicated that the archetypical approaches could be transferred to the new domain but the related domain data is consistently sub par when compared to high quality in-domain data. Finally, the viability of the best classification algorithm could be qualitatively verified in a real-world setting that was also developed within the course of this thesis.