This thesis documents the grammatical processes and semantic impact of innovative ways to pejoratively reference individuals through adjectival nominalization. Research on nominalized adjectives suggests that when meanings shift from having one property (1) to becoming a kind with associated properties (2), the noun form often encodes stereotypical attributes: [1] "Her hair is blonde." (hair color); [2] "He married a blonde." (female, sexy, dumb). Likewise, the linguistic phenomenon of genericity refers to classes or kinds and different grammatical structures reflect properties in different ways. In 1 and 2 above, the shift from adjectival blonde to indefinite NP a blonde moves the focus from the definitional characteristic to the prototypical. Similarly, adjectival gay [3] is definitional, but the marked, nominal form [4] adds socially-based conceptions of the "average" gay (example from Twitter): [3] jesus christ i make a joke and now im a gay man? (sexuality) [constructed]; [4] jesus christ i make a joke and now im a gay? … (flamboyant, abnormal). To investigate innovative reference via nominalization, two corpus studies based in human judgment were conducted. In the first study, a subset of the corpus (N=121) was annotated for pejoration by five additional linguists following the same guidelines as the original annotator. In the second study, 800 instances were annotated by non-experts using crowd-sourcing. In both studies we find a correspondence between nominal status and pejorative meaning.

評論是所有人類行為的核心，因為它影響我們行為的關鍵因素。我們都試著從不同型式的評論分析與研究試著從作者字裡行間的文字呈現內容深入推敲及理解，從而要能過濾出能協助決策的有用資訊。在早期的評論研究將評論視為是文本分類問題，直到2000年前後，從分析評論的主觀句子與評論裡形容詞的程度衡量用詞，學者們開始對解構整篇文本的內容，並試著從語言學的角度分析用字遣詞與情感方向之間的關聯。這種從文字語義關聯分析評論的方式，也使文本挖掘技術必需結合自然語言的處理原則，才能更準確地了解評論的內容。隨著許多新興的機器學習演算法與自然語言處理方法不斷地推陳出新，及網路使用行為拓展至電子商務與線上虛擬社群的建立，情感分析研究亦開始不斷地蓬勃發展。 漢文不同於世界其它語言，它擁有許多獨特表徵：無空格區隔、一字一語素、依詞為語言中表達意義的最小獨立單位，也使得在套用源自西方的情感分析原則時更加困難。然而過去的研究者則加以利用這些語言特徵，建立出專屬中文的情感分析原則。我們務實地討論適用於中文情感分析的情境(a)可取得情感分析資源及專家語言智慧，及(b)可取得領域字詞特徵向量定義的兩個前題下，提出適合的中文情感分析策略。在情境(a)中，我們深入討論運用部首資訊至情感分析中的適用性，並且提出一套能精萃出領域評論文本的觀測字詞/部首組的方法。研究中我們萃取出50個部首組，並運用在領域相近的評論裡得到很好的情感分類成效。而在情境(b)中我們提出適合深層類神經網路學習方法的評論字詞的權重過濾原則，不僅能確保評論字詞在學習過程中仍保有能積旋出合適屬性，並且驗證此權重原則在支援向量機的學習方式下亦有相同的優勢。在研究中，我們亦討論此兩種情境下進行情感分析的必要條件與資訊，並為未來更深入的中文情感分析起到墊腳石的作用。 / Opinion is the core of human behaviors, because it directly influences key factor of our behaviors. Despite of personal or organizational decision making processes, we all constantly conduct various kinds of opinion analysis, including explaining and comprehending what users present. At the beginning, opinion studies considered as a text mining problems, and tried to cluster opinions into positive and negative groups. After 2000, researchers intended to decompose sentences from whole opinions by analysing subjective expressing and adjective words presenting within, as well as explained the relationships between semantics and sentiment from linguistics aspect. Therefore, opinion analysis has to incorporate with natural language processing techniques, so we can understand the opinion contents. Nowadays, sentiment analysis grows event booming due to emerging machine learning and natural language processing approaches, as well as the needs of electronic commerce and virtual community on line. Unfortunately, Chinese is quite unlike other language due to non-space separated, one character as one morpheme, and considering words (compositing with several characters) as minimum semantic expression unit. And those language features also bring difficult to adopted sentiment analysis principles from English. Nevertheless, researchers leveraged Chinese language information to propose specific sentiment analysis approaches dedicated to analyze Chinese opinions. In this study, we practically discussed the situations of conducting sentiment analysis: (a) using sentiment analysis resources and experts’ knowledge; and (b) using word feature vector, called word2vec, and deep learning. In (a) scenario, we propose a Chinese radical-based sentiment analysis approach and experiment the applicability. We also proposed a feature extraction method, so we can generate 50 seeds for further analysis. In (b), we compared 4 different feature selection approaches for deep learning, in order to keep accuracy and make sure understandable feature can be generated in neural network. We also tested feature selection approaches in SVM classifier and retrieved similar results. In this study, we also discussed essential constraints and required information in both scenarios, as well as the results of this study can be the foundation of continuing Chinese sentiment analysis studies.

近年來傳統線上購物正面臨著一連串的市場困境，如削價競爭、廉價品競爭等，因此導致銷售量之成長趨緩，反觀跨境線上購物卻出現了蓬勃發展的態勢，因而讓跨境線上購物成為驅動經濟活動與國際貿易的新引擎。另一方面，由於跨境線上購物的情境複雜性遠高於傳統的境內線上購物，業者們欲開發一海外新市場，必須先了解該地消費者行為與其購買決策過程後，才能制定出好的商業策略，並且進一步將產品導向的服務轉化成為以顧客導向的服務，才有機會為傳統線上購物之困境另闢生機。因此，引取並了解消費者所體認的內在價值是經營跨境線上購物最重要的成功因素。 本研究將試圖將傳統境內線上購物研究擴展到跨境線上購物議題，藉由文字探勘(Text Mining)分析、語意情感分析與 k-means 分群演算法，挖掘出消費者對於所購買商品之評論的常見內容型態與所購買商品之類別，並試圖找出跨境網購平台上各項因素及商品評論對於產品銷售量間之關連性，提供未來研究者及跨境網購平台業者決策之依據。 / While online shopping websites are facing the difficulties of price and low-quality competition, cross-border online shopping is on a vigorous development trend, showing that cross-border online shopping is an important trend of online shopping field. Due to the complexity of cross-border online shopping is much higher than the traditional domestic online shopping, so understanding the value of cross-border online shopping consumers is the most important success factors. Companies want to develop new markets abroad, must understand the local consumer’s behaviour and their decision-making process in order to make good business strategies. This study uses text mining analytic technology, semantic analysis techniques, and k-means clustering algorithm to identify characteristics of consumers’ reviews and the common categories of goods they purchased. After getting the reason why consumers use cross-border online shopping service and what values they got in this service. Researcher can predict and analyse the evolution and development of cross-border online shopping, provide reference for future online shopping academic studies and online shopping industry’s decision-making.

近來Instagram成為流行的分享照片社交平台。在上傳影像到網路社交平台時，人們透過套用不同的濾鏡來表達他們的感受。然而，對於修改過的影像，我們不太可能逆向推回得知影像套用了什麼樣的濾鏡。本研究嘗試透過定義出十種影像風格，對應於一些最常應用的濾鏡，來解決這種逆向工程問題。因此，原始問題被轉化為分類問題，並可以使用機器學習方法來解決。為了生成訓練數據，我們根據用戶投票收集標記的結果。根據我們的實驗，在調查中概述的十個類別中，投票的結果有很高的共識。我們在HSV空間中使用分析出的顏色特徵來區分影像風格，並採用支持向量機（SVM）做分類。驗證我們數據集中的Top 1和Top 3準確度分別為64%和96%，顯示機器分類的效能與人類觀察者的效能相當。最後，我們導入數位著名攝影師的作品，進行個案研究，以測試風格識別和情感分析結果。 / Recently, Instagram has become a very popular social media platform for sharing photos. People apply different type of filters to express their feelings when posting photos on social networking sites. Given a filtered image, it is difficult, if not possible, to determine which filter has been applied to obtain the observed effects. This study attempts to address this reverse engineering problem by defining ten image styles corresponding to some of the most frequently applied filters. As such, the original question is cast into a classification problem which can be solved using machine learning approaches. To generate training data, we collected the labeled results based on user votes. Consensuses among users are found to be high in the ten categories outlined in our investigation. We employ color features in the HSV space to characterize image styles. Support vector machine (SVM) is then used for classification. The accuracies for top-1 and top-3 category using our dataset are 64% and 96%, respectively. The performance of machine classification is comparable to that of human observers. Finally, works by famous photographers are brought in to validate the style recognition and sentiment analysis results.

Contemporary approaches to natural language processing are predominantly based on statistical machine learning from large amounts of text, which has been manually annotated with the linguistic structure of interest. However, such complete supervision is currently only available for the world's major languages, in a limited number of domains and for a limited range of tasks. As an alternative, this dissertation considers methods for linguistic structure prediction that can make use of incomplete and cross-lingual supervision, with the prospect of making linguistic processing tools more widely available at a lower cost. An overarching theme of this work is the use of structured discriminative latent variable models for learning with indirect and ambiguous supervision; as instantiated, these models admit rich model features while retaining efficient learning and inference properties. The first contribution to this end is a latent-variable model for fine-grained sentiment analysis with coarse-grained indirect supervision. The second is a model for cross-lingual word-cluster induction and the application thereof to cross-lingual model transfer. The third is a method for adapting multi-source discriminative cross-lingual transfer models to target languages, by means of typologically informed selective parameter sharing. The fourth is an ambiguity-aware self- and ensemble-training algorithm, which is applied to target language adaptation and relexicalization of delexicalized cross-lingual transfer parsers. The fifth is a set of sequence-labeling models that combine constraints at the level of tokens and types, and an instantiation of these models for part-of-speech tagging with incomplete cross-lingual and crowdsourced supervision. In addition to these contributions, comprehensive overviews are provided of structured prediction with no or incomplete supervision, as well as of learning in the multilingual and cross-lingual settings. Through careful empirical evaluation, it is established that the proposed methods can be used to create substantially more accurate tools for linguistic processing, compared to both unsupervised methods and to recently proposed cross-lingual methods. The empirical support for this claim is particularly strong in the latter case; our models for syntactic dependency parsing and part-of-speech tagging achieve the hitherto best published results for a wide number of target languages, in the setting where no annotated training data is available in the target language.

Social media today is a dominant communication tool, which structures not only our social interactions but also filter the information users are getting displayed. The big social media platforms use our interaction data to analyse our behaviour and sell the data for commercial interest. But not only the pure interaction data is valuable for these platforms. Also hidden information, which can be derived from our interactive networks, about our social structures, social classifications and social status are gathered and monetised. This research attempts on the one hand to uncover some of these methods used by social media platforms, and on the other hand, also wants to show how useful these new methods can be for research on social phenomena. Therefore, this study goes beyond the confining limits of traditional sociology, where either qualitative or quantitative methods are applied. Following the idea of Critical Realism, the positivist and constructivist methods are applied in combination in order to provide thick accounts of the studied material. In this study, varying socioeconomic classification systems (like the Sinus-Milieu models) are investigated and evaluated against the background of Bourdieu’s ideas on cultural and social forms of capital. The present study uses a mixed method approach (Social Network Analysis and Sentiment Analysis) to analyse quantitative data from Twitter conversations which were collected during the Austrian EU Election 2019. In conclusion, one could say that the overall purpose of this study is to demonstrate the usefulness of Critical Realism for social media research, since this approach can create a thicker account of the studied material than other, more traditional methods. This undertaking is demonstrated by the findings of the study. These findings are the building of specific sub-clusters of EU candidates which are not related to the same political background and traditional demographics but whose relation can be detected and described using Bourdieu’s concepts of social and cultural capital. As a mean for gathering empirical data, Twitter turned out to be a useful and accessible tool for this study.

Extraire l'opinion publique en analysant les Big Social data a connu un essor considérable en raison de leur nature interactive, en temps réel. En effet, les données issues des réseaux sociaux sont étroitement liées à la vie personnelle que l’on peut utiliser pour accompagner les grands événements en suivant le comportement des personnes. C’est donc dans ce contexte que nous nous intéressons particulièrement aux méthodes d’analyse du Big data. La problématique qui se pose est que ces données sont tellement volumineuses et hétérogènes qu’elles en deviennent difficiles à gérer avec les outils classiques. Pour faire face aux défis du Big data, de nouveaux outils ont émergés. Cependant, il est souvent difficile de choisir la solution adéquate, car la vaste liste des outils disponibles change continuellement. Pour cela, nous avons fourni une étude comparative actualisée des différents outils utilisés pour extraire l'information stratégique du Big Data et les mapper aux différents besoins de traitement.La contribution principale de la thèse de doctorat est de proposer une approche d’analyse générique pour détecter de façon automatique des tendances d’opinion sur des sujets donnés à partir des réseaux sociaux. En effet, étant donné un très petit ensemble de hashtags annotés manuellement, l’approche proposée transfère l'information du sentiment connue des hashtags à des mots individuels. La ressource lexicale qui en résulte est un lexique de polarité à grande échelle dont l'efficacité est mesurée par rapport à différentes tâches de l’analyse de sentiment. La comparaison de notre méthode avec différents paradigmes dans la littérature confirme l'impact bénéfique de notre méthode dans la conception des systèmes d’analyse de sentiments très précis. En effet, notre modèle est capable d'atteindre une précision globale de 90,21%, dépassant largement les modèles de référence actuels sur l'analyse du sentiment des réseaux sociaux. / Extracting public opinion by analyzing Big Social data has grown substantially due to its interactive nature, in real time. In fact, our actions on social media generate digital traces that are closely related to our personal lives and can be used to accompany major events by analysing peoples' behavior. It is in this context that we are particularly interested in Big Data analysis methods. The volume of these daily-generated traces increases exponentially creating massive loads of information, known as big data. Such important volume of information cannot be stored nor dealt with using the conventional tools, and so new tools have emerged to help us cope with the big data challenges. For this, the aim of the first part of this manuscript is to go through the pros and cons of these tools, compare their respective performances and highlight some of its interrelated applications such as health, marketing and politics. Also, we introduce the general context of big data, Hadoop and its different distributions. We provide a comprehensive overview of big data tools and their related applications.The main contribution of this PHD thesis is to propose a generic analysis approach to automatically detect trends on given topics from big social data. Indeed, given a very small set of manually annotated hashtags, the proposed approach transfers information from hashtags known sentiments (positive or negative) to individual words. The resulting lexical resource is a large-scale lexicon of polarity whose efficiency is measured against different tasks of sentiment analysis. The comparison of our method with different paradigms in literature confirms the impact of our method to design accurate sentiment analysis systems. Indeed, our model reaches an overall accuracy of 90.21%, significantly exceeding the current models on social sentiment analysis.

L’objectif de cette thèse est d’étudier les sentiments dans les documents comparables. Premièrement, nous avons recueillis des corpus comparables en anglais, français et arabe de Wikipédia et d’Euronews, et nous avons aligné ces corpus au niveau document. Nous avons en plus collecté des documents d’informations des agences de presse locales et étrangères dans les langues anglaise et arabe. Les documents en anglais ont été recueillis du site de la BBC, ceux en arabe du site d’Al-Jazzera. Deuxièmement, nous avons présenté une mesure de similarité cross-linguistique des documents dans le but de récupérer et aligner automatiquement les documents comparables. Ensuite, nous avons proposé une méthode d’annotation cross-linguistique en termes de sentiments, afin d’étiqueter les documents source et cible avec des sentiments. Enfin, nous avons utilisé des mesures statistiques pour comparer l’accord des sentiments entre les documents comparables source et cible. Les méthodes présentées dans cette thèse ne dépendent pas d’une paire de langue bien déterminée, elles peuvent être appliquées sur toute autre couple de langue / The aim of this thesis is to study sentiments in comparable documents. First, we collect English, French and Arabic comparable corpora from Wikipedia and Euronews, and we align each corpus at the document level. We further gather English-Arabic news documents from local and foreign news agencies. The English documents are collected from BBC website and the Arabic documents are collected from Al-jazeera website. Second, we present a cross-lingual document similarity measure to automatically retrieve and align comparable documents. Then, we propose a cross-lingual sentiment annotation method to label source and target documents with sentiments. Finally, we use statistical measures to compare the agreement of sentiments in the source and the target pair of the comparable documents. The methods presented in this thesis are language independent and they can be applied on any language pair

O mercado cinematográfico pode ser caracterizado como uma indústria de entretenimento com a produção de bens de informação que são também bens de experiência, cuja qualidade só é conhecida após o consumo. Deste modo, a revisão crítica se torna importante para induzir seu consumo, fornecendo previamente algum grau de informação sobre a qualidade do bem. Segue-se o trabalho de Reinstein e Snyder (2005) para determinar se as revisões críticas conduzidas por consumidores e por críticos profissionais online afetam o tempo de exibição de filmes no mercado americano de cinema, medido em quantidades de semanas, conforme modelos de duração/sobrevivência na literatura. Para esta finalidade foi gerado, a partir de sites de cinemas americanos (Box Office Mojo e Rotten Tomatoes), um banco de dados extremamente rico com informações semanais de todos os filmes disponíveis no cinema americano de 2004 a 2015. Especificamente, investigou-se os efeitos das revisões críticas de críticos profissionais de primeira linha (Tops) e de consumidores, conforme a média das notas atribuídas na semana de lançamento de cada filme. No que se refere à avaliação dos consumidores foi aplicada a computação afetiva, que reconhece o sentimento e a emoção em suas resenhas online para captar o efeito boca a boca potencializado pelas mídias sociais e fornecendo, portanto, uma análise mais profunda do boca a boca online. O estudo controla por possíveis problemas de endogeneidade decorrente de simultaneidade, usando as críticas somente antes e durante a semana de lançamento dos filmes. Os resultados sugerem que os críticos profissionais exercem grande influência no tempo de duração dos filmes em cartaz, bem como a positividade dos consumidores em relação ao filme. No entanto, o efeito dos críticos profissionais é em média 3 vezes maior do que dos consumidores. Adicionalmente, pode-se observar que algumas emoções afetam a expectativa de vida dos filmes a depender do gênero do mesmo / The movie market may be considered as entertainment industry, which produces experience goods that is also information goods, whose quality is only known only after consumption. Thus, critical reviews becomes important to induce consumption, since it provides some level of information about product quality. We follow Reinstein and Snyder (2005) works in order to determine if experts and consumers online critical reviews affect the survival time of movies at the American movie market, measured by number of weeks, according to survival analysis models in the literature. For this purpose, an extremely rich database with weekly information on all the films available in American cinema from 2004 to 2015 was generated from American movie sites (Box Office Mojo and Rotten Tomatoes). Specifically, we investigate the effects of critical reviews from top professionals and from consumers, according to the average ratings assigned in each movie\'s release week. As far as consumer assessment was concerned, affective computing was applied, which recognizes the sentiment (sentiment analysis) and emotion (emotion mining) in their online reviews to capture the word-of-mouth effect boosted by social media. The study controls for possible problems of endogeneity due to simultaneity, using the criticisms before and during the week of release of the films. The results suggest that the professional critics exert a great influence on the duration of the films in exhibition, as well as the positivity of the consumers in relation to the film. Thus, the effect of professionals are 5 times greater, generally, than the effect of the consumer critics. Additionally, it can be observed that some emotions affect movie life expectancy depending on the its genre

Submitted by Fabricia Fialho Reginato (fabriciar) on 2015-07-04T00:45:28Z No. of bitstreams: 1 SilviaFoschiera.pdf: 3755161 bytes, checksum: 0f631548f2054c557658d1a50094a5ac (MD5) / Made available in DSpace on 2015-07-04T00:45:28Z (GMT). No. of bitstreams: 1 SilviaFoschiera.pdf: 3755161 bytes, checksum: 0f631548f2054c557658d1a50094a5ac (MD5) Previous issue date: 2012-07-31 / Nenhuma / O objetivo principal desta investigação é verificar em que aspectos a Semântica de Frames (FILLMORE, 1982; 1985) e o modelo denominado Roda das Emoções (SCHERER, 2005) contribuem na relação entre a linguagem e o fenômeno da emoção, considerando os idiomas português e espanhol. A Semântica de Frames, perspectiva teórica vinculada à Linguística Cognitiva, fundamenta a análise semântica e sintática por meio de um estudo exploratório do maquinário da FrameNet (FILLMORE et al., 2003). Com base nesse arcabouço teórico, realizamos um levantamento dos frames e elementos de frame de verbos e adjetivos que descrevem a emoção, associando categorias semânticas e sintáticas. Verificamos, também, a possibilidade de mapear o holder e o tópico de opinião considerando o corpus de sentenças do Twitter. A segunda perspectiva teórica está relacionada à Psicologia Cognitiva, por meio do modelo denominado Roda das Emoções. Considerando os traços semânticos sugeridos nessa ferramenta, observa-se o quanto, levando em conta aplicações computacionais, ela vem enriquecer um estudo de Análise de Sentimento. A Roda das Emoções serve para identificar a polaridade das opiniões constantes por meio dos adjetivos nas sentenças da amostra. Os resultados evidenciam que as duas perspectivas se revelam produtivas para aplicações computacionais em Análise de Sentimento. / The main objective of this research is to ascertain which aspects of Frame Semantics (Fillmore, 1982; 1985) and the model called Wheel of Emotions (Scherer, 2005) contribute to the relationship between language and the phenomenon of emotion, in regards to the Portuguese and Spanish languages. Frame Semantics -a theoretical construct linked to cognitive linguistics- underlies the syntactic and semantic analysis by means of an exploratory study of the FrameNet database (Fillmore et al., 2003). Based on this theoretical framework, we conducted a survey of the frames and frame elements that describe emotions, attaching semantic and syntactic categories to them. We also contemplated the possibility of mapping the holder and the subject of opinion in the corpus of sentences from Twitter. The second theoretical perspective is related to cognitive psychology through the Wheel of Emotions. Considering the semantic aspects offered by this tool, it is observed to what extent –including computer applications- it enriches the study of Sentiment Analysis. The Wheel of Emotions helps to identify the polarity of opinions contained in the sample sentences. The results show that the two perspectives prove productive for computer applications in Sentiment Analysis.