<div> <div> <p>Benefited from the booming social network, reading posts from other users overinternet is becoming one of commonest ways for people to intake information. Onemay also have noticed that sometimes we tend to focus on users provide well-foundedanalysis, rather than those merely who vent their emotions. This thesis aims atfinding a simple and efficient way to recognize reliable information sources amongcountless internet users by examining the sentiments from their past posts.<br></p><p>To achieve this goal, the research utilized a dataset of tweets about Apples stockprice retrieved from Twitter. Key features we studied include post-date, user name,number of followers of that user, and the sentiment of that tweet. Prior to makingfurther use of the dataset, tweets from users who do not have sufficient posts arefiltered out. To compare user sentiments and the derivative of Apples stock price, weuse Pearson correlation between them for to describe how well each user performs.Then we iteratively increase the weight of reliable users and lower the weight ofuntrustworthy users, the correlation between overall sentiment and the derivative ofstock price will finally converge. The final correlations for individual users are theirperformance scores. Due to the chaos of real world data, manual segmentation viadata visualization is also proposed as a denoise method to improve performance.Besides our method, other metrics can also be considered as user trust index, suchas numbers of followers of each user. Experiments are conducted to prove that ourmethod out performs others. With simple input, this method can be applied on awide range of topics including election, economy, and job market.<br></p> </div> </div>

O projeto descrito neste documento propõe um modelo para análise de viés em notícias, procurando identificar o viés dos meios de comunicação em relação a entidades políticas. Foram analisados três tipos de viés: o viés de seleção, que avalia o quanto uma entidade é referenciada pelo meio de comunicação; o viés de cobertura, que avalia quanto destaque é destinado a entidade e, por fim, o viés de afirmação, que avalia se estão falando mal ou bem da entidade. Para tal, foi construído um corpus de notícias sistematicamente extraídas de 5 produtores de notícias e classificadas manualmente em relação à polaridade e entidade alvo. Técnicas de análise de sentimentos baseadas em aprendizado de máquina foram validadas utilizando o corpus criado. Criou-se uma metodologia para identificação de viés, utilizando o conceito de outliers, a partir de métricas indicadoras. A partir da metodologia proposta, foi analisado o viés em relação aos candidatos ao governo de São Paulo e à presidência a partir do corpus criado, em que se identificou os três tipos de viés em dois produtores de notícias / The project described here proposes a model to study bias on newswire texts, related to political entities. Three types of bias are analysed: selection bias, which refers to the amount of times an entity is referenced by the media outlet; coverage bias, which assesses the amount of coverage given to an entity and, finally, the assertion bias, which analyses whether the news is a positive or negative report of an entity. To accomplish this, a corpus was systematically built by extracting news from 5 different newswires. These texts were manually classified according to their polarity alignment and associated entity. Sentiment Analysis techniques were applied and evaluated using the corpus. Based on the concept of outliers, a methodology for bias detection was created. Bias was analysed using the proposed methodology on the generated corpus for candidates to the government of the state of São Paulo and to presidency, being identified in two newswires for the three above-defined types

A large proportion of World Wide Web (WWW) users treat it as a social medium, i.e. many of them use the WWW to express and communicate their opinions. Economic value or utility can be created if these utterances, reactions, or feedback are extracted from various social media platforms and their content analysed. Some of these benefits are related to e-commerce, marketing, product improvements, improving machine learning algorithms etc. Moreover, establishing links between different social media platforms, based on shared topics and content, could provide access to the comments of users of different platforms. However, studies to date have generally tackled the area of content extraction from each type of social media in isolation. There is a lack of research of some aspects of social media, namely, linking the references from a blog post, for example, to information related to the same issue on Twitter. In addition, while studies have been carried out on various languages, there has been little investigation into social media in the Arabic language. This thesis tackles opinion mining and sentiment analysis of Arabic language social media, particularly in blogs and Twitter. The thesis focuses on Arabic language technology blogs in order to identify the expressed sentiments and then to link an issue within a blog post to relevant tweets in Twitter. This was done by assessing the similarity of content and measuring the sentiments scores. In order to extract the required data, text-mining techniques were used to build up corpora of the raw blog data in Modern Standard Arabic (MSA) and to build tools and lexicons required for this research. The results obtained through this research contribute to the field of computer science by furthering the employment of text-mining techniques, thus improving the process of information retrieval and knowledge accumulation. Moreover, the study developed new approaches to working with Arabic opinion mining and the domain of sentiment analysis.

O projeto descrito neste documento propõe um modelo para análise de viés em notícias, procurando identificar o viés dos meios de comunicação em relação a entidades políticas. Foram analisados três tipos de viés: o viés de seleção, que avalia o quanto uma entidade é referenciada pelo meio de comunicação; o viés de cobertura, que avalia quanto destaque é destinado a entidade e, por fim, o viés de afirmação, que avalia se estão falando mal ou bem da entidade. Para tal, foi construído um corpus de notícias sistematicamente extraídas de 5 produtores de notícias e classificadas manualmente em relação à polaridade e entidade alvo. Técnicas de análise de sentimentos baseadas em aprendizado de máquina foram validadas utilizando o corpus criado. Criou-se uma metodologia para identificação de viés, utilizando o conceito de outliers, a partir de métricas indicadoras. A partir da metodologia proposta, foi analisado o viés em relação aos candidatos ao governo de São Paulo e à presidência a partir do corpus criado, em que se identificou os três tipos de viés em dois produtores de notícias / The project described here proposes a model to study bias on newswire texts, related to political entities. Three types of bias are analysed: selection bias, which refers to the amount of times an entity is referenced by the media outlet; coverage bias, which assesses the amount of coverage given to an entity and, finally, the assertion bias, which analyses whether the news is a positive or negative report of an entity. To accomplish this, a corpus was systematically built by extracting news from 5 different newswires. These texts were manually classified according to their polarity alignment and associated entity. Sentiment Analysis techniques were applied and evaluated using the corpus. Based on the concept of outliers, a methodology for bias detection was created. Bias was analysed using the proposed methodology on the generated corpus for candidates to the government of the state of São Paulo and to presidency, being identified in two newswires for the three above-defined types

Energy has been at the top of the national and global political agenda along with other

abstract: Internet sites that support user-generated content, so-called Web 2.0, have become part of the fabric of everyday life in technologically advanced nations. Users collectively spend billions of hours consuming and creating content on social networking sites, weblogs (blogs), and various other types of sites in the United States and around the world. Given the fundamentally emotional nature of humans and the amount of emotional content that appears in Web 2.0 content, it is important to understand how such websites can affect the emotions of users. This work attempts to determine whether emotion spreads through an online social network (OSN). To this end, a method is devised that employs a model based on a general threshold diffusion model as a classifier to predict the propagation of emotion between users and their friends in an OSN by way of mood-labeled blog entries. The model generalizes existing information diffusion models in that the state machine representation of a node is generalized from being binary to having n-states in order to support n class labels necessary to model emotional contagion. In the absence of ground truth, the prediction accuracy of the model is benchmarked with a baseline method that predicts the majority label of a user's emotion label distribution. The model significantly outperforms the baseline method in terms of prediction accuracy. The experimental results make a strong case for the existence of emotional contagion in OSNs in spite of possible alternative arguments such confounding influence and homophily, since these alternatives are likely to have negligible effect in a large dataset or simply do not apply to the domain of human emotions. A hybrid manual/automated method to map mood-labeled blog entries to a set of emotion labels is also presented, which enables the application of the model to a large set (approximately 900K) of blog entries from LiveJournal. / Dissertation/Thesis / M.S. Computer Science 2011

abstract: Muslim radicalism is recognized as one of the greatest security threats for the United States and the rest of the world. Use of force to eliminate specific radical entities is ineffective in containing radicalism as a whole. There is a need to understand the origin, ideologies and behavior of Radical and Counter-Radical organizations and how they shape up over a period of time. Recognizing and supporting counter-radical organizations is one of the most important steps towards impeding radical organizations. A lot of research has already been done to categorize and recognize organizations, to understand their behavior, their interactions with other organizations, their target demographics and the area of influence. We have a huge amount of information which is a result of the research done over these topics. This thesis provides a powerful and interactive way to navigate through all this information, using a Visualization Dashboard. The dashboard makes it easier for Social Scientists, Policy Analysts, Military and other personnel to visualize an organization's propensity towards violence and radicalism. It also tracks the peaking religious, political and socio-economic markers, their target demographics and locations. A powerful search interface with parametric search helps in narrowing down to specific scenarios and view the corresponding information related to the organizations. This tool helps to identify moderate Counter-Radical organizations and also has the potential of predicting the orientation of various organizations based on the current information. / Dissertation/Thesis / M.S. Computer Science 2012

Ingeniero Civil en Computación / La masividad del uso de las redes sociales ha crecido explosivamente en los últimos años. Resulta interesante conocer la opinión que expresan los usuarios en Twitter para realizar estudios de mercado, popularidad de marcas, candidatos presidenciales, etc. Este trabajo tiene por objetivo desarrollar un software que permita hacer análisis de opinión en Twitter. Este software se utilizó para estudiar la opinión sobre los candidatos a presidente en el año 2013 en Chile. Se estudiaron dos técnicas utilizadas para obtener el sentimiento asociado a un texto: Método Estadístico y Método Ontológico. El primer método requiere de un gran volumen de datos (textos de los que se conoce si expresan una opinión positiva o negativa) para entrenar el algoritmo. Se eligió el método ontológico, para el que se construyen manualmente reglas para identificar el sentimiento. Para aplicar estas reglas, se procesa el texto libre usando la librería FreeLing, que construye un árbol de dependencia de las palabras que componen el texto. Dicho árbol permite agrupar el sujeto con los correspondientes adjetivos, verbos, etc de las oraciones. La ontología construida consiste en patrones detectables en los arboles de dependencia, con palabras claves que pueden ir en las distintas posiciones del patrón. Hubo problemas con la librería FreeLing que no procesa correctamente texto mal escrito, como es el caso típico de los Tweets. Se tuvo que hacer un preprocesamiento al texto para ayudar a FreeLing a procesar el texto. Al hacer el análisis de los Tweets de los 7 días anteriores a la segunda vuelta, se obtuvo una popularidad del 61% para Bachelet (obtuvo 62% en las elecciones) y un 39% para Matthei (38% en las elecciones), resultado que también es cercano a las estimaciones de Brandmetrics. Otra funcionalidad desarrollada es la identificación de la posición geográfica del usuario, y por lo tanto sus Tweets, a partir del dato que él indica en el perfil de usuario. Este es un campo de texto libre. El texto se trata de calzar con una serie de expresiones regulares, que están asociadas con las regiones de Chile. Se validó la técnica desarrollada comparando los resultados obtenidos con los datos por GPS para aquellos Tweets para los que estaban disponibles, obteniendo cerca de un 90% de acierto. Sin embargo, sólo a alrededor de la mitad de los Tweets se les puede identificar la localidad usando esta técnica, lo que de todas formas es mejor que cerca del 2% de los Tweets que tienen la información del GPS. Se analizó también el uso de Twitter en función de la hora del día, observando la máxima actividad en la noche, durante y después de los noticieros.

Ingeniero Civil Industrial / Los contenidos disponibles en la Web están creciendo a velocidades que hacen que la tarea de analizarlos sea humanamente imposible. Una de las disciplinas que hace frente a este problema es la Minería de Opiniones, también conocida como el Análisis de Sentimientos, responsable de procesar texto automáticamente, con el fin de extraer y analizar las opiniones que contiene para generar información valiosa y accionable. El objetivo principal de este trabajo es crear una aplicación de Minería de Opiniones capaz de explotar tweets en español que mencionen a la empresa de retail Falabella. En primer lugar, se investigó el impacto que las redes sociales tienen en Chile. En segundo lugar, se elaboró un estado del arte que englobara los últimos avances en Minería de Opiniones y en Procesamiento del Lenguaje Natural. En tercer lugar, se creó un Web Crawler capaz de obtener los tweets que mencionanaran a la compañía. Posteriormente se implementó varios algoritmos de Procesamiento del Lenguaje Natural para pre-procesar los tweets previamente mencionados, e incorporar los datos resultantes al proceso de extracción de opiniones. Este proceso se desarrolló como un enfoque de Minería de Opiniones no supervisado basado en lexicones, dependiente de un analizador de dependencias encargado de detectar ciertas estructuras gramaticales que permitieran identificar fenómenos linguísticos comunes, tales como la negación, intensificación, y oraciones subordinadas adversativas. La identificación de dichos fenómenos permitió mejorar la calidad de la clasificación. Finalmente se creó una página Web para mostrar los resultados que luego fueron utilizados para realizar un análisis exploratorio de la compañía. Adicionalmente, los algoritmos fueron validados con el corpus TASS, obteniendo valores-F de un 61,88% negativo y 71,88% positivo. A pesar de que el rendimiento de los algoritmos no fue tan alto como una aplicación en producción lo requeriría, se consideró lo suficientemente bueno como para realizar el análisis exploratorio. Con éste fue posible confirmar la intuición de que las cuentas corporativas suelen publicar contenido positivo, las cuentas de noticias contenido neutral, y los usuarios comunes contenido irrelevante o quejas. Además fue posible probar que los usuarios más activos frecuentemente publican contenido totalmente irrelevante. Por otra parte, se logró replicar varios resultados obtenidos por instituciones nacionales reconocidas, entre los cuales destaca el hecho que el momento más controversial del año para Falabella fue cuando se intentó llevar a cabo el Cyber Monday, período en el cual el sentimiento generalizado en Twitter alcanzó los niveles más negativos. Dicho todo esto, la aplicación desarrollada demostró ser útil al momento de utilizar una gran cantidad de datos para extraer información que podría ser potencialmente útil para la firma de retail. Finalmente, el desarrollo de la aplicación permitió crear un artículo que contuviera parte considerable del transfondo teórico en el cual ésta se basó, además de beneficiar a otros estudiantes en el desarrollo de sus memorias.

In recent years, increasing attentions are paid on developing exceptional technologies for efficiently processing massive collection of heterogeneous data generated by different kinds of sensors. While we have observed great successes of utilizing big data in many innovative applications, the need on integrating information poses new challenges caused by the heterogeneity of the data. In this thesis, we target at geo-tagged data, and propose a cloud based platform named City Digital Pulse (CDP), where a unified mechanism and extensible architecture are provided to facilitate the various aspects in big data analysis, ranging from data acquisition to data visualization. We instantiate the proposed system using multi-model data collected from two social platforms, Twitter and Instagram, which include plenty of geo-tagged messages. Data analysis is performed to detect human affections from the user uploaded content. The emotional information in big social data can be uncovered by using a multi-dimension visualization interface, based on which users can easily grasp the evolving of human affective status within a given geographical area, and interact with the system. This offers costless opportunities to improve the decision making in many critical areas. Both the proposed architecture and algorithm are empirically demonstrated to be able to achieve real-time big data analysis.