Dissertations / Theses on the topic 'Understanding of the text'

Dissertação (mestrado) - Universidade Federal de Santa Catarina, Centro de Comunicação e Expressão. Programa de Pós-Graduação em Letras/Inglês e Literatura Correspondente.<br>Made available in DSpace on 2012-10-20T16:03:13Z (GMT). No. of bitstreams: 1 210520.pdf: 6154986 bytes, checksum: 19acef4a425b924af72b8f876e19cbd6 (MD5)<br>Estudo das relações texto-imagem em artigos de capa da revista Newsweek, visando contribuir para o entendimento de como significados multimodais são construídos. A partir da macro-análise de 24 artigos de capa, identificam-se os principais componentes verbais e visuais da estrutura deste gênero multimodal. Enquanto que a partir da micro-análise de dois artigos de capa, investiga-se como os modos verbal e visual constroem significados funcionais e como estes significados modulam, daí construindo o significado central dos artigos de capa. Baseado nos resultados da investigação proposta, o estudo aponta para três urgentes necessidades pedagógicas.

Dissertação (mestrado) - Universidade Federal de Santa Catarina. Centro de Comunicação e Expressão<br>Made available in DSpace on 2016-01-08T16:51:35Z (GMT). No. of bitstreams: 1 80067.pdf: 4788847 bytes, checksum: e2967ec153e31fb0d4401ad3f98eadc2 (MD5) Previous issue date: 1990

En aquesta tesi s'aborda el problema de la comprensió de text automàtic en condicions sense restriccions. En particular, abordem les tasques de detecció, el seguiment, i el reconeixement de text multi-idioma amb orientació arbitrària en imatges i vídeos d'escena natural. Per a això s'ha desenvolupat un conjunt de mètodes genèrics que es basen de la premissa bàsica que el text té sempre algunes característiques claus visuals que són independents de la llengua o escriptura en el qual està escrit. Les metodologies d'extracció de text en escena es basen generalment en la classificació de regions o regions individuals, utilitzant el coneixement a priori d'un script o llenguatge determinat. La percepció humana de text, per contra, es basa en l'organització perceptiva a través del qual el text apareix com un grup perceptualment rellevant d'objectes atòmics. En aquesta tesi, s'argumenta que el problema de l'extracció de text pot ser plantejat com a detecció de grups significatius de regions de la imatge. Ens ocupem del problema de la segmentació del text en escenes naturals des d'una perspectiva jeràrquica, fent ús explícit de l'estructura del text, enfocant directament la detecció d'agrupacions de regions que corresponen a elements de text dins d'una jerarquia produïda per un procés d'aglomeració per similitud sobre les regions individuals. Proposem una manera òptima per construir aquesta jerarquia de manera que s'estableix un espai de característiques dissenyat per produir hipòtesis de grup de text, i una regla de detenció que combina un classificador discriminatiu i una mesura probabilística de rellevancia dels grups amb base en l'organització perceptiva. Proposem un nou algoritme de propostes d'objecte que està dissenyat específicament per a detectar text, i el comparem amb altres mètodes genèrics en l'estat de la tècnica. Alhora, s'estudia en quina mesura els mètodes existents de propostes d'objectes genèrics poden ser útils per a la detecció de text d'escena. A continuació, presentem un algoritme híbrid per a la detecció i seguiment de text d'escena en el qual la noció de groups de regions també juga paper central. Un mòdul d'extracció de text escena basada en MSER ​​s'utilitza per detectar el text de forma asíncrona, mentre que paral·lelament els objectes de text detectates són seguits per propagació de MSER. La cooperació d'aquests dos mòduls va més enllà del seguiment per detecció en termes d'optimització de temps, i pot oferir processament de vídeo en temps real fins i tot en dispositius de baixos recursos. Finalment, ens centrem en el problema de la identificació de l'escriptura en imatges de text d'escena per tal de construir un sistema de lectura d'extrem a extrem en escenris multi-idioma. Afrontar aquest problema amb classificadors CNN no és una tasca senzilla, ja que aquests no tenen en compte una característica clau de les instàncies de text d'escena: la seva relació d'aspecte es extremadament variable. En lloc de canviar la mida d'imatges d'entrada a una mida fixa, com en l'ús típic dels classificadors CNN holístics, proposem un marc de classificació basat en caracteristiques locals amb la finalitat de preservar les parts discriminatives de la imatge que són característiques de la seva classe. Es descriu un nou mètode basat en l'ús de conjunts de xarxes unides per aprendre conjuntament representacions discriminatives de regions locals de la imatge i alhora la seva importància relativa en un esquema de classificació global. Els nostres experiments amb aquest procediment d'aprenentatge demostren la viabilitat d'identificació d'alfabet en imatges d'escena natural, aplanant el camí cap als sistemes de comprensió de text d'escena d'extrem a extrem en diversos idiomes.<br>En esta tesis se aborda el problema de la comprensión de texto automático en condiciones sin restricciones. En particular, abordamos las tareas de detección, seguimiento, y el reconocimiento de texto multi-idioma con orientación arbitraria en imágenes y vídeos de escena natural. Para ello se ha desarrollado un conjunto de métodos genéricos que se basan de la premisa básica de que el texto tiene siempre algunas características claves visuales que son independientes de la lengua o escritura en el que está escrito. Las metodologías de extracción de texto en escena se basan generalmente en la clasificación de regiones o regiones individuales, utilizando el conocimiento a priori de un script o lenguaje determinado. La percepción humana de texto, por el contrario, se basa en la organización perceptiva a través del cual el texto aparece como un grupo perceptualmente relevante de objetos atómicos. En esta tesis, se argumenta que el problema de la extracción de texto puede ser planteado como detección grupos significativos de regiones de la imagen. Nos ocupamos del problema de la segmentación del texto en escenas naturales desde una perspectiva jerárquica, haciendo uso explícito de la estructura del texto, enfocando directamente la detección de agrupaciones de regiones que corresponden a elementos de texto dentro de una jerarquía producida por un proceso de aglomeración por similitud sobre las regiones individuales. Proponemos una manera óptima para construir esta jerarquía de modo que se establece un espacio de características diseñado para producir hipótesis de grupo de texto, y una regla de detención que combina un clasificador discriminativo y una medida probabilística de relevancia de los grupos con base en el organización perceptiva. Proponemos un nuevo algoritmo de propuestas de objeto que está diseñado específicamente para detectar texto, y lo comparamos con otros métodos genéricos en el estado de la técnica. Asimismo, se estudia en qué medida los métodos existentes de propuestas de objetos genéricos pueden ser útiles para la detección de texto de escena. A continuación, presentamos un algoritmo híbrido para la detección y seguimiento de texto de escena en el que la noción de groups de regiones también juega papel central. Un módulo de extracción de texto escena basada en MSER utiliza para detectar el texto de forma asíncrona, mientras que paralelamente los objetos de texto detectados son seguidos por propagación de MSER. La cooperación de estos dos módulos va más allá del seguimiento para detección en términos de optimización de tiempo, y puede ofrecer procesamiento de vídeo en tiempo real incluso en dispositivos de bajos recursos. Finalmente, nos centramos en el problema de la identificación de la escritura en imágenes de texto de escena para construir un sistema de lectura de extremo a extremo en escenris multi-idioma. Afrontar este problema con clasificadores CNN no es una tarea sencilla, ya que estos no tienen en cuenta una característica clave de las instancias de texto de escena: su relación de aspecto es extremadamente variable. En lugar de cambiar el tamaño de imágenes de entrada a un tamaño fijo, como en el uso típico de los clasificadores CNN holísticos, proponemos un marco de clasificación basado en características locales con el fin de preservar las partes discriminativas de la imagen que son características de su clase. Se describe un nuevo método basado en el uso de conjuntos de redes unidas para aprender conjuntamente representaciones discriminativas de regiones locales de la imagen y al mismo tiempo su importancia relativa en un esquema de clasificación global. Nuestros experimentos con este procedimiento de aprendizaje demuestran la viabilidad de identificación de alfabeto en imágenes de escena natural, allanando el camino hacia los sistemas de comprensión de texto de escena de extremo a extremo en varios idiomas.<br>This thesis addresses the problem of automatic scene text understanding in unconstrained conditions. In particular, we tackle the tasks of multi-language and arbitrary-oriented text detection, tracking, and recognition in natural scene images and videos. For this we have developed a set of generic methods that build on top of the basic assumption that text has always some visual key characteristics that are independent of the language or script in which it is written. Scene text extraction methodologies are usually based in classification of individual regions or patches, using a priori knowledge for a given script or language. Human perception of text, on the other hand, is based on perceptual organisation through which text emerges as a perceptually significant group of atomic objects. In this thesis, we argue that the text extraction problem could be posed as the detection of meaningful groups of regions. We address the problem of text segmentation in natural scenes from a hierarchical perspective, making explicit use of text structure, aiming directly to the detection of region groupings corresponding to text within a hierarchy produced by an agglomerative similarity clustering process over individual regions. We propose an optimal way to construct such an hierarchy introducing a feature space designed to produce text group hypotheses with high recall and a novel stopping rule combining a discriminative classifier and a probabilistic measure of group meaningfulness based in perceptual organization. We propose a new Object Proposals algorithm that is specifically designed for text and compare it with other generic methods in the state of the art. At the same time we study to what extent the existing generic Object Proposals methods may be useful for scene text understanding. Then, we present a hybrid algorithm for detection and tracking of scene text where the notion of region grouppings plays also central role. A scene text extraction module based on Maximally Stable Extremal Regions (MSER) is used to detect text asynchronously, while in parallel detected text objects are tracked by MSER propagation. The cooperation of these two modules goes beyond the full-detection approaches in terms of time performance optimization, and yields real-time video processing at high frame rates even on low-resource devices. Finally, we focus on the problem of script identification in scene text images in order to build a multi-language end-to-end reading system. Facing this problem with state of the art CNN classifiers is not straightforward, as they fail to address a key characteristic of scene text instances: their extremely variable aspect ratio. Instead of resizing input images to a fixed size as in the typical use of holistic CNN classifiers, we propose a patch-based classification framework in order to preserve discriminative parts of the image that are characteristic of its class. We describe a novel method based on the use of ensembles of conjoined networks to jointly learn discriminative stroke-parts representations and their relative importance in a patch-based classification scheme. Our experiments with this learning procedure demonstrate the viability of script identification in natural scene images, paving the road towards true multi-lingual end-to-end scene text understanding.

The obesity and overweight growing rates are a major global concern for both developed and developing countries. Digital technologies can potentially deliver effective interventions to tackle this crisis, with mobile phones uniquely positioned to deliver scalable, real-time, inexpensive, and interactive support for all at risk. A meta-analysis showed that SMS interventions are effective for promoting weight loss and physical activity. The content of SMS messages of digital behaviour change interventions is usually seen as the main driver of effectiveness. It has been suggested that personalised communication, reaching people on emotional as well as rational levels, is more effective. However, there is no previous study exploring people’s affective responses to message wordings and personalisation, which could lead to more effective tools for promoting healthier behaviours, like physical activity. Affect is a term used to define the experience of a feeling or an emotion. Psychophysiological and self-reported measures of arousal and valence, the two main dimensions of emotions, were used to measure message impact on different levels of consciousness; and as methods to better understand people’s response to affective messages and personalisation. The first study in this thesis examined electrodermal activity and facial electromyography as objective measures of arousal and valence, respectively, and found strong associations between these measures. Study 2 compared psychophysiological and self-reported responses to previously evaluated affective and cognitive messages but found no significant difference between them. Study 3 used the same methods to test the effect of personalisation of affective and cognitive messages using identification (participant’s name) and contextualisation (participant’s preferred physical activity). Personalisation of messages using contextualisation and identification were found to be effective strategies for eliciting emotional responses and persuasiveness. This thesis contributes to our knowledge about using psychophysiology and self-report as methods to measure people’s response to affective and personalised messages, and the value of measuring the emotional impact of text messages on the recipient before using these in randomised controlled trials. The developers of SMS interventions and other digital techniques could benefit from using these methods. Future work needs to investigate the impact of messages designed using these methods on actual behaviour.

Context, Text and Reader is an investigation of the way that members of a young audience use magazine and television advertisements to help them develop an understanding of technology and technological development. The thesis' main argument is that understanding the relationship between context, text and reader is the key to understanding how an audience uses advertising images to help it come to terms with the world. A combination of methodologies is employed, including semiotic analysis, content analysis, attitude surveys and group discussions. The study is divided into four main sections. The first, Context, provides a framework into which the empirical findings can be loosely fitted. Two important areas are looked at: the changes taking place in what is often referred to as the communications revolution and secondly, the role that advertising plays in the generation of social and cultural meaning. The second section, Text, includes a semiotic analysis of selected television and magazine advertisements for technology products. This is followed by a review of a number of published content analysis studies and then by a discussion of two empirical studies of television and magazine advertisements using a content analysis methodology. The third section moves on to investigate the reader's response to technology. The findings of three empirical investigations are outlined and discussed. These are an exploration of attitudes to, firstly, technology in general; secondly, a television commercial for a domestic personal computer and thirdly a commercial for a television set. The final section of the thesis, Context, Text and Reader, brings together the three different elements of the study with, first of all, an outline of the issues involved in what are called "Reception Studies", then by an analysis of two group discussions of the television commercials for a domestic personal computer and a television set, used earlier in the study.

This thesis addresses the problem of text spotting - being able to automatically detect and recognise text in natural images. Developing text spotting systems, systems capable of reading and therefore better interpreting the visual world, is a challenging but wildly useful task to solve. We approach this problem by drawing on the successful developments in machine learning, in particular deep learning and neural networks, to present advancements using these data-driven methods. Deep learning based models, consisting of millions of trainable parameters, require a lot of data to train effectively. To meet the requirements of these data hungry algorithms, we present two methods of automatically generating extra training data without any additional human interaction. The first crawls a photo sharing website and uses a weakly-supervised existing text spotting system to harvest new data. The second is a synthetic data generation engine, capable of generating unlimited amounts of realistic looking text images, that can be solely relied upon for training text recognition models. While we define these new datasets, all our methods are also evaluated on standard public benchmark datasets. We develop two approaches to text spotting: character-centric and word-centric. In the character-centric approach, multiple character classifier models are developed, reinforcing each other through a feature sharing framework. These character models are used to generate text saliency maps to drive detection, and convolved with detection regions to enable text recognition, producing an end-to-end system with state-of-the-art performance. For the second, higher-level, word-centric approach to text spotting, weak detection models are constructed to find potential instances of words in images, which are subsequently refined and adjusted with a classifier and deep coordinate regressor. A whole word image recognition model recognises words from a huge dictionary of 90k words using classification, resulting in previously unattainable levels of accuracy. The resulting end-to-end text spotting pipeline advances the state of the art significantly and is applied to large scale video search. While dictionary based text recognition is useful and powerful, the need for unconstrained text recognition still prevails. We develop a two-part model for text recognition, with the complementary parts combined in a graphical model and trained using a structured output learning framework adapted to deep learning. The trained recognition model is capable of accurately recognising unseen and completely random text. Finally, we make a general contribution to improve the efficiency of convolutional neural networks. Our low-rank approximation schemes can be utilised to greatly reduce the number of computations required for inference. These are applied to various existing models, resulting in real-world speedups with negligible loss in predictive power.

In this thesis we consider the problems that Proper Names cause in the analysis of unedited, naturally-occurring text. Proper Names cause problems because of their high frequency in many types of text, their poor coverage in conventional dictionaries, their importance in the text understanding process, and the complexity of their structure and the structure of the text which describes them. For the most part these problems have been ignored in the field of Natural Language Processing, with the result that Proper Names are one of its most under-researched areas. As a solution to the problem, we present a detailed description of the syntax and semantics of seven major classes of Proper Name, and of their surrounding context. This description leads to the construction of syntactic and semantic rules specifically for the analysis of Proper Names, which capitalise on the wealth of descriptive material which often accompanies a Proper Name when it occurs in a text. Such an approach side-steps the problem of lexical coverage, by allowing a text processing system to use the very text it is analysing to construct lexical and knowledge base entries for unknown Proper Names as it encounters them. The information acquired on unknown Proper Names goes considerably beyond a simple syntactic and semantic classification, instead consisting of a detailed genus and differentia description. A complete solution to the 'Proper Name Problem' must include approaches to the handling of apposition, conjunction and ellipsis, abbreviated reference, and many of the far from standard phenomena encountered in naturally-occurring text. The thesis advances partial and practical solutions in all of these areas. In order to set the work described in a suitable context, the problems of Proper Names are viewed as a subset of the general problem of lexical inadequacy, as it arises in processing real, un-edited, text. The whole of this field is reviewed, and various methods of lexical acquisition compared and evaluated. Our approach to coping with lexical inadequacy and to handling Proper Names is implemented in a news text understanding system called FUNES, which is able to automatically acquire detailed genus and differentia information on Proper Names as it encounters them in its processing of news text. We present an assessment of the system's performance on a sample of unseen news text which is held to support the validity of our approach to handling Proper Names.

This thesis describes a study that was carried out in response to the question, "how do I encourage active reading behaviour among a group of less successful L2 readers?" I focus on talk around text as an alternative to the traditional classroom reading practices. I chose action research approach because it provided a context for achieving the following aims: to gain an understanding of the L2 readers' difficulties in constructing understanding; to implement and evaluate talk around text; and to reflect on my own teaching practice with a view to creating opportunities for students' engagement in meaningful reading activities. Throughout the study, I monitored the intervention activities and made amendments for the subsequent cycles by examining the data available from observational field notes, audiorecorded reading activities, student evaluation of the activities (by means of interviews, informal conversations, written comments, questionnaire) and the evaluation meetings with my critical friends. Data analysis showed that the unsuccessful L2 readers were able to solve problems and engage in rich discussions about what they had read when they were provided with the opportunity to talk around text. Further, implementing change meant that I had to examine and overcome the contradictions that existed between my belief about learning and my classroom practice. In this thesis, I discuss my attempts to overcome the contradictions and I critically reflect on the opportunities and challenges associated with my attempt to change. Future research activities should consider issues beyond classroom practice that may inhibit teachers and students from exercising change. Implementing change on classroom practice alone will not be sufficient without considering ways to influence change in the wider educational community.

The concept 'mess-understanding' has circulated in online media and is so prevalent that it is now included in the Urban Dictionary. The folk concept of mess-understandings is a pun for misunderstandings arising in an online media context. Posting one's own or others' miscommunication and/or typographical errors has grown to be a popular way of sharing humor via cross-platform sharing on the Internet. The aim of this paper is to analyze short message service (SMS) dialogues shared in social media, with a special emphasis on those with the highest degree of 'shareability' and/or popularity. The study specifically focuses on understanding linguistic and communicative reasons behind these dialogues being treated as humorous by users. As such, the study aims to shed light upon current cultural conceptions of communication and humor. Data was collected from the photo sharing website Pinterest from users who had posted or reposted 'screen shots' from their own or others' SMS conversations. In order to collect as much valuable data as possible, a manual search strategy was developed with three different word strings which resulted in a corpus of 160 dialogues. Content analysis of the data revealed certain recurrent humor themes, such as allusions to sexual conduct or bodily functions, generation gaps, technology difficulties, and lexical ambiguity.<br>Begreppet ‘mess-förstånd’ har på senare tid cirkulerat på Internet och är nu så allmänt förekommande att det är inkluderat i Urban Dictionary. Mess-förstånd är en ordvits för missförstånd som förekommer på Internet. Att lägga upp sina egna eller andras misslyckade konversationer och/eller typografiska fel har utvecklats till ett populärt sätt att dela humor via olika plattformar på Internet. Syftet med den här uppsatsen är att analysera vilka lingvistiska mönster av SMS-missförstånd som finns, med speciell betoning på sådana som har den högsta graden av ‘delbarhet’ och/eller popularitet och vad det säger oss om rådande föreställningar om kommunikation och humor. Det empiriska materialet insamlades från fotodelningshemsidan Pinterest från användare som hade lagt upp eller återbrukat ‘skärmavbilder’ från sina egna eller andras SMS-konversationer. För att samla in så mycket värdefull data som möjligt användes en manuell sökstrategi med tre olika ordsträngar som resulterade i en korpus med 160 dialoger. Analyser av dialogerna visade på återkommande humorteman, exempelvis anspelningar på sexualitet eller kroppsliga funktioner, generationsklyftor, teknologiska svårigheter och lexikala tvetydligheter.

Cette thèse travaille à identifier ce qu’un modèle de reconnaissance de texte apprend pendant son entraînement, à travers l’examen du contenu de ses vérités de terrain et de ses erreurs de prédiction. L’intention principale ici est d’améliorer les connaissances sur le fonctionnement d’un réseau de neurones, avec des expériences focalisées sur des documents tapuscrits. Les méthodes utilisées se sont concentrées surtout sur l’exploration approfondie des données d’entraînement, l’observation des erreurs de prédiction des modèles et la corrélation entre les deux. Une première hypothèse, basée sur l’influence du lexique, fut non concluante. Cependant, cela a dirigé les observations vers un nouveau niveau d’étude, s’appuyant sur un niveau infralexical : les n-grammes. La distribution de ceux des données d’entraînement a été analysée et subséquemment, comparée à celle des n-grammes récupérés dans les erreurs de prédiction. Des résultats prometteurs ont conduit à une exploration approfondie, tout en passant d’un modèle de langue unique à un modèle multilingue. Des résultats concluants m’ont permis de déduire que les n-grammes pourraient effectivement être une réponse valide aux performances de reconnaissance<br>This thesis works on identifying what a text recognition model can learn during its training, through the examination of its ground truth’s content, and its prediction’s errors. The main intent here is to improve the knowledge of how a neural network operates, with experiments focused on typewritten documents. The methods used mostly concentrated on the thorough exploration of the training data, the observation of the model’s prediction’s errors, and the correlation between both. A first hypothesis, based on the influence of the lexicon, was inconclusive. However, it steered the observation towards a new level of study, relying on an infralexical level: the n-grams. Their training data’s distribution was analysed and subsequently compared to that of the n-grams retrieved from the prediction errors. Promising results lead to further exploration, while upgrading from single-language to multilingual model. Conclusive results enabled me to infer that the n-grams might indeed be a valid answer to recognition’s performances

Book Summary: New Ways in Teaching With Music shows how music can be incorporated into your lessons as a way to decrease anxiety, increase motivation and retention, and invigorate both students and teachers. This book is a collection of adaptable lessons that use music as a catalyst for effective, engaging, and enjoyable language learning. 101 activities for students of all skill levels and a companion website with online resources are included. The lessons are broken down by topic including: Reading, Writing, Listening, Speaking, Grammar, Vocabulary, Cultural Exploration, Integrated Skills

Event-related collections, including both tweets and webpages, have valuable information, and are worth exploring in interdisciplinary research and education. Unfortunately, such data is noisy, so this variety of information has not been adequately exploited. Further, for better understanding, more knowledge hidden behind events needs to be unearthed. Regarding these collections, different societies may have different requirements in particular scenarios. Some may need relatively clean datasets for data exploration and data mining. Social researchers require preprocessing of information, so they can conduct analyses. General societies are interested in the overall descriptions of events. However, few systems, tools, or methods exist to support the flexible use of event-related collections. In this research, we propose a new, integrated system to process and analyze event-related collections at different levels (i.e., data, information, and knowledge). It also provides various services and covers the most important stages in a system pipeline, including collection development, curation, analysis, integration, and visualization. Firstly, we propose a query likelihood model with pre-query design and post-query expansion to rank a webpage corpus by query generation probability, and retrieve relevant webpages from event-related tweet collections. We further preserve webpage data into WARC files and enrich original tweets with webpages in JSON format. As an application of data management, we conduct an empirical study of the embedded URLs in tweets based on collection development and data curation techniques. Secondly, we develop TwiRole, an integrated model for 3-way user classification on Twitter, which detects brand-related, female-related, and male-related tweeters through multiple features with both machine learning (i.e., random forest classifier) and deep learning (i.e., an 18-layer ResNet) techniques. As guidance to user-centered social research at the information level, we combine TwiRole with a pre-trained recurrent neural network-based emotion detection model, and carry out tweeting pattern analyses on disaster-related collections. Finally, we propose a tweet-guided multi-document summarization (TMDS) model, which generates summaries of the event-related collections by using tweets associated with those events. The TMDS model also considers three aspects of named entities (i.e., importance, relatedness, and diversity) as well as topics, to score sentences in webpages, and then rank selected relevant sentences in proper order for summarization. The entire system is realized using many technologies, such as collection development, natural language processing, machine learning, and deep learning. For each part, comprehensive evaluations are carried out, that confirm the effectiveness and accuracy of our proposed approaches. Regarding broader impact, the outcomes proposed in our study can be easily adopted or extended for further event analyses and service development.<br>Doctor of Philosophy<br>Event-related collections, including both tweets and webpages, have valuable information. They are worth exploring in interdisciplinary research and education. Unfortunately, such data is noisy. Many tweets and webpages are not relevant to the events. This leads to difficulties during data analysis of the datasets, as well as explanation of the results. Further, for better understanding, more knowledge hidden behind events needs to be unearthed. Regarding these collections, different groups of people may have different requirements. Some may need relatively clean datasets for data exploration. Some require preprocessing of information, so they can conduct analyses, e.g., based on tweeter type or content topic. General societies are interested in the overall descriptions of events. However, few systems, tools, or methods exist to support the flexible use of event-related collections. Accordingly, we describe our new framework and integrated system to process and analyze event-related collections. It provides varied services and covers the most important stages in a system pipeline. It has sub-systems to clean, manage, analyze, integrate, and visualize event-related collections. It takes an event-related tweet collection as input and generates an event-related webpage corpus by leveraging Wikipedia and the URLs embedded in tweets. It also combines and enriches original tweets with webpages. As an application of data management, we conduct an empirical study of tweets and their embedded URLs. We developed TwiRole for 3-way user classification on Twitter. It detects brand-related, female-related, and male-related tweeters through their profiles, tweets, and images. To aid user-centered social research, we combine TwiRole with an existing emotion detection tool, and carry out tweeting pattern analyses on disaster-related collections. Finally, we propose a tweet-guided multi-document summarization (TMDS) model and service, which generates summaries of the event-related collections by using tweets associated with those events. It extracts important sentences across different topics from webpages, and organizes them in proper order. The entire system is realized using many technologies, such as collection development, natural language processing, machine learning, and deep learning. For each part, comprehensive evaluations help confirm the effectiveness and accuracy of our proposed approaches. Regarding broader impact, our methods and system can be easily adopted or extended for further event analyses and service development.

<p>Mitt syfte med undersökningen är att konstruera en modell för bedömning av elevtexter. Bedömningsmodellen ska vara kapabel att ge både en helhetsbild och en detaljbild av elevers diskursiva analytiska förmåga. Min utgångspunkt vid konstruktionen av bedömningsmodellen var bland annat Judith Langers föreställningsvärld och John Biggs SOLO-taxonomi. För att testa min bedömningsmodell samlade jag in elevtexter från naturvetarprogrammet på en gymnasieskola i Södertälje kommun. Texterna är inhämtade från två ämnen, Svenska A och Svenska C, och består av essäer. Resultatet av min undersökning blev en bedömningsmatris som granskar texterna utifrån fyra aspekter; struktur, språk och stil, argumentation och reflektion. Bedömningsmatrisen fungerar emellertid bättre som ett vertyg för uppskattning av texter än som ett vertyg för bedömning. Den ger tillräcklig information hur eleverna ligger till enligt kursplanerna för Svenska A respektive C inom dessa fyra aspekter, och utifrån de resultaten kan läraren se vad varje enskild elev klarar av och inte klarar av och kan planera lektionerna därefter.</p>

For decades electrophysiologists have recorded and characterized the biophysical properties of a rich diversity of neuron types. This diversity of neuron types is critical for generating functionally important patterns of brain activity and implementing neural computations. In this thesis, I developed computational methods towards quantifying neuron diversity and applied these methods for understanding the functional implications of within-type neuron variability and across-type neuron diversity. First, I developed a means for defining the functional role of differences among neurons of the same type. Namely, I adapted statistical neuron models, termed generalized linear models, to precisely capture how the membranes of individual olfactory bulb mitral cells transform afferent stimuli to spiking responses. I then used computational simulations to construct virtual populations of biophysically variable mitral cells to study the functional implications of within-type neuron variability. I demonstrate that an intermediate amount of intrinsic variability enhances coding of noisy afferent stimuli by groups of biophysically variable mitral cells. These results suggest that within-type neuron variability, long considered to be a disadvantageous consequence of biological imprecision, may serve a functional role in the brain. Second, I developed a methodology for quantifying the rich electrophysiological diversity across the majority of the neuron types throughout the mammalian brain. Using semi-automated text-mining, I built a database, Neuro- Electro, of neuron type specific biophysical properties extracted from the primary research literature. This data is available at http://neuroelectro.org, which provides a publicly accessible interface where this information can be viewed. Though the extracted physiological data is highly variable across studies, I demonstrate that knowledge of article-specific experimental conditions can significantly explain the observed variance. By applying simple analyses to the dataset, I find that there exist 5-7 major neuron super-classes which segregate on the basis of known functional roles. Moreover, by integrating the NeuroElectro dataset with brain-wide gene expression data from the Allen Brain Atlas, I show that biophysically-based neuron classes correlate highly with patterns of gene expression among voltage gated ion channels and neurotransmitters. Furthermore, this work lays the conceptual and methodological foundations for substantially enhanced data sharing in neurophysiological investigations in the future.

The explosive growth in the number of documents produced daily necessitates the development of effective alternatives to explore, analyze, and discover knowledge from documents. Document mining research work has emerged to devise automated means to discover and analyze useful information from documents. This work has been mainly concerned with constructing text representation models, developing distance measures to estimate similarities between documents, and utilizing that in mining processes such as document clustering, document classification, information retrieval, information filtering, and information extraction. <br /><br /> Conventional text representation methodologies consider documents as bags of words and ignore the meanings and ideas their authors want to convey. It is this deficiency that causes similarity measures to fail to perceive contextual similarity of text passages due to the variation of the words the passages contain, or at least perceive contextually dissimilar text passages as being similar because of the resemblance of words the passages have. <br /><br /> This thesis presents a new paradigm for mining documents by exploiting semantic information of their texts. A formal semantic representation of linguistic inputs is introduced and utilized to build a semantic representation scheme for documents. The representation scheme is constructed through accumulation of syntactic and semantic analysis outputs. A new distance measure is developed to determine the similarities between contents of documents. The measure is based on inexact matching of attributed trees. It involves the computation of all distinct similarity common sub-trees, and can be computed efficiently. It is believed that the proposed representation scheme along with the proposed similarity measure will enable more effective document mining processes. <br /><br /> The proposed techniques to mine documents were implemented as vital components in a mining system. A case study of semantic document clustering is presented to demonstrate the working and the efficacy of the framework. Experimental work is reported, and its results are presented and analyzed.

Detta examensarbete syftar till att undersöka vilka strategier elever använder sig av vid text-, hör- och ordförståelse i franska. Jag har genom undervisningsförsök och elevenkäter undersökt vilka strategier elever i en niondeklass har när de arbetar med ovanstående moment där uppgifterna är hämtade utanför den ordinarie läroboken. Jag har därutöver intervjuat lärare på den skola jag gjort mina undervisningsförsök för att se om de ger eleverna strategier för arbete kring text-, hör- och ordförståelse. Projektet STRIMS och Tornbergs forskningsresultat har givit mig inspiration till arbetet. Eleverna använde sig huvudsakligen av framgångsrika strategier när de tog sig an de olika uppgifterna. De framgångsrika strategier eleverna i huvudsak använde sig av var att de drog slutsatser utifrån kontexten, de använde sina kunskaper i andra språk samt att de utbytte tankar och idéer vid samarbete i par eller i grupp.

To help librarians understand the impact of emerging technologies on participatory service building, this study takes social media as an example to explore how to use different engagement strategies that social media provides to engage more users. This research provides three major contributions to the library system. The libraries can use the resultant engagement strategies to engage its users. Additionally, the best-fit strategy can be inferred and designed based on users' preferences. Lastly, the users' preferences can be understood based on data analysis of social media. Three such contributions have been put together to fully address the proposed question.

The aim of the present study was to examine students' text comprehension in grades 4 and 6. Participants were 61 students without documented or suspected reading and writing difficulties, in grades 4 and 6. The focus was on texts designed by Bishop and Adams (1992), containing two types of questions: literal and inference questions, and three different ways of presenting the text; listening, silent reading and silent reading with the text available when responding to the questions. The texts have been previously used in research and as a tool for clinical, qualitative estimate of reading comprehension in speech and language pathologists’ assessments. Investigation was also conducted for working memory, vocabulary, and phonological and orthographical decoding. The results show that students have a significantly higher score in terms of literal questions when the text is available in responding to the questions, versus listening and silent reading. The results also show that literal and inference questions do not differ. No differences between gender and grades were encountered. Conclusions are drawn that students with typical reading development in grades 4 and 6 do not have any difficulty in answering neither literal nor inference questions, concerning the three texts in this study.

The research is an exploratory study of reader-text relationship with a primary focus on reader knowledge and use of text structure in order to gain a further insight into expository reading practice at an advanced level of second language learning. The study was carried out with 165 Maltese learners of English at the post-secondary level oflanguage learning engaged in preparation for academic study at university.

The purpose of this study was to understand the differences of conscious constructive responses and shifting strategies with the goal of understanding reading comprehension while reading nonfiction, fiction, and poetry texts. This naturalistic diagnostic study uses a think-aloud methodology. The study examines verbal, retrospective, and recall reports from fifteen seventh-graders of varying reading abilities; good, average, and weaker from one private school in the Midwest. After collecting a total of ninety verbal reports in three sessions, verbal reports were scored according to level of conscious constructive responses. Retrospective reports were scored for shifting utilization. Recall reports were scored to identify a reading comprehension score that was correlated with strategies. Based on these results, it was evident that conscious constructive responses existed with seventh-grade readers regardless of ability and text type. There were differences between utilization with text type, but little differences with ability level. Shifting strategies were evident with seventh graders, but there were no significant differences when text type and ability level were taken into consideration. Two conscious constructive responses, relating text to text and relating text to prior knowledge, correlated with recall scores demonstrating an increase in reading comprehension. Two shifting strategies, making liberal interpretations and looking for useful information significantly improved reading comprehension. Findings support the model of constructively responsive reading (Pressley & Afflerbach, 1995).<br>Department of Elementary Education

This thesis explores the effects of emotional information about characters in text on processing. In five chapters presenting nine experiments in total, readers were presented with emotional characters that occurred either in small texts or in sentences. In the second chapter, it was investigated whether mental representations of entities in sentences are more salient and easier to retrieve due to emotional information. In the third chapter, the effects of emotional information about multiple different characters on processing were explored. The forth chapter presents experiments on perspective taking and how perspective affects the way emotional information is processed. Building up on that, in chapter 5, it was investigated how the mood of the reader influences perspective taking when reading about emotional information. All experiments in the first four chapters used a self-paced reading method to explore effects on reading speed (reading times). Chapter 6, however, presents an eye-tracking experiment set out to explore the effects of perspective on reading behaviour in more detail and to determine where perspective differences arise in the text. Hence, pronoun regions (including perspective cues) across the text were analysed. The findings presented in this thesis gave evidence that readers focus more on emotional characters (that emotional characters are more salient), and that readers also engage more with (emotional) texts when they experience the situation from a personal perspective. All experiments gave evidence that readers track and use emotional information to form a coherent representation of the text.

This thesis focuses on how Graduate Diploma Primary pre-service teachers experience the process of becoming the teacher. It argues that this graduate cohort of teachers come to teacher education with previous experiences of teachers and teaching, life histories, dispositions and a complex personal life that inform the process of becoming. Understanding the teacher self, leading to the development of an identity as the teacher and the dispositions that enable the individual to cope with the difficulties and complexities that becoming the teacher entails are a focus of the research. The context of the study is a school of education within a mid sized Australian University. The study is informed by a social constructivist theoretical approach coupled with a literacy model borrowed from the work of two literacy theorists. A framework for reading and comprehending the text of teaching is developed to enable an understanding of the process of becoming the teacher. Four case studies are developed to explicate the story of each individual utilizing three interviews throughout the two semesters of their program. Drawings of themselves as the teacher they envisioned themselves to be and a graph of their year provided rich data enabling a telling of each story. The research builds a picture of the experiences of four graduate diploma pre-service teachers as they negotiate the development of an identity as the teacher. The findings indicate that becoming the teacher is a deeply personal process undertaken in the contexts of others. It is complex and difficult work particularly for mature aged pre-service teachers who have much in their lives other than their teacher education program. It is about reading the text of teaching and not only developing understandings about the text but also being able to transform that understanding into new ways of working within a school and classroom. A teacher education program that emphasizes the personal and that has a collaborative and collegiate connection with schools is proposed in order to enhance the possibilities for successful outcomes for pre-service teachers.

Summarization models which operate on meaning representations of documents have been neglected in the past, although they are a very promising and interesting class of methods for summarization and text understanding. In this thesis, I present one such summarizer, which uses the proposition as its meaning representation. My summarizer is an implementation of Kintsch and van Dijk's model of comprehension, which uses a tree of propositions to represent the working memory. The input document is processed incrementally in iterations. In each iteration, new propositions are connected to the tree under the principle of local coherence, and then a forgetting mechanism is applied so that only a few important propositions are retained in the tree for the next iteration. A summary can be generated using the propositions which are frequently retained. Originally, this model was only played through by hand by its inventors using human-created propositions. In this work, I turned it into a fully automatic model using current NLP technologies. First, I create propositions by obtaining and then transforming a syntactic parse. Second, I have devised algorithms to numerically evaluate alternative ways of adding a new proposition, as well as to predict necessary changes in the tree. Third, I compared different methods of modelling local coherence, including coreference resolution, distributional similarity, and lexical chains. In the first group of experiments, my summarizer realizes summary propositions by sentence extraction. These experiments show that my summarizer outperforms several state-of-the-art summarizers. The second group of experiments concerns abstractive generation from propositions, which is a collaborative project. I have investigated the option of compressing extracted sentences, but generation from propositions has been shown to provide better information packaging.

Electronic word-of-mouth (eWOM) has increasingly become an important topic in marketing and consumer research. However, theory construction and methodology development in this area are still in their infancy. This leaves some basic and important questions unanswered including whether eWOM communication is effective, what roles are played by different communication cues, and how valuable information from text reviews can be generated. This study intends to answer these fundamental questions.Based on the Brunswik's Len Model, this study developed the Process Model of eWOM Communication. It extends the Brunswik's Lens Model in several important ways and provides a systematic tool to examine the effectiveness of eWOM communication processes. Furthermore, a simplified model of eWOM communication was developed to test the validity of automatic text analysis as a promising tool in studying eWOM communication.Two focus group interviews and a throughout literature review were conducted first to identify the communication cues employed by eWOM partners. Then, two web-based self-administered surveys were carried out to collect data from both eWOM senders and readers. Last, the data from both eWOM senders and readers were matched, forming a final dataset with 90 reviews. Correlations, regressions, and path analyses were employed to evaluate the models and test the hypotheses.Results showed that eWOM communication is effective, and the relative strength of information flow varies in different eWOM communication links when communicating different types of information.This study identified a list of eWOM communication cues and found that consumers employ different cues in communicating different types of information. EWOM readers' inference structure in decoding may not exactly mirror eWOM senders' encoding structure. Moreover, communication cues especially verbal cues play an important role in eWOM communication and explain additional variance in eWOM partners' intentions and perceptions beyond and above the star ratings. In general, negative emotion words are the most important cues across various situations.In addition, this study provides initial evidence for the validity of automatic text analysis in studying eWOM. Linguistic indicators such as Negations, Negative Emotions, and Money can explain additional variance in eWOM partners' attitudes and emotions beyond and above the star ratings.

Despite the enormous body of literature studying the typing errors of adults, children's typing errors remain an understudied area. It is well known in the field of Child-Computer Interaction that children are not 'little adults'. This means findings regarding how adults make typing mistakes cannot simply be transferred into how children make typing errors, without first understanding the differences. To understand how children differ from adults in the way they make typing mistakes, typing data were gathered from both children and adults. It was important that the data collected from the contrasting participant groups were comparable. Various methods of collecting typing data from adults were reviewed for suitability with children. Several issues were identified that could create a bias towards the adults. To resolve these issues, new tools and methods were designed, such as a new phrase set, a new data collector and new computer experience questionnaires. Additionally, there was a lack of an analysis method of typing data suitable for use with both children and adults. A new categorisation method was defined based on typing errors made by both children and adults. This categorisation method was then adapted into a Java program, which dramatically reduced the time required to carry out typing categorisation. Finally, in a large study, typing data collected from 231 primary school children, aged between 7 and 10 years, and 229 undergraduate computing students were analysed. Grouping the typing errors according to the context in which they occurred allowed for a much more detailed analysis than was possible with error rates. The analysis showed children have a set of errors they made frequently that adults rarely made. These errors that are specific to children suggest that differences exist between the ways the two groups make typing errors. This finding means that children's typing errors should be studied in their own right.

This qualitative study explores how women in emerging adulthood make meaning of text messaging while developing an intimate heterosexual relationship. A focus group consisting of eight women was used in order to understand the way women interpret text messages during the formation of a romantic relationship. Data were analyzed using thematic analysis. Four themes emerged from the data: (1) what is the landscape?, (2) how do I figure it out?, (3) red flags, and (4) a red flag means. Participants described the context in which they enter relationships, and the role texting plays. Different strategies they used to make meaning of ambiguous text messages were discussed. Participants reported behaviors that were identified as red flags and how participants made meaning of those red flags Limitations, clinical implications, and future directions are identified and discussed.<br>Master of Science

[Abstract]: This study focuses on the nature, role and function of online social presence in text-based online learning environments and how it relates to learner support. Ultimately, the study seeks to understand the relationships between the nature, role and function of social presence, the social learning processes employed in contemporary online learning environments and the mechanisms which support learning in these environments. This study is guided by the question: In what ways might social presence enhance the provision of learner support in text-based online learning environments? Of particular interest in this study is information related to (a) how participants in online courses experience social presence; (b) how learners in online courses perceive and/or experience learner support in online courses and (c) how these experiences affect learner activity in online learning environments.The research design employs a collective case study approach which focuses on learners in four separate online courses within a single university context to inform understanding of the phenomena in question. Using a combination of interview-like techniques, the study seeks to ground the understanding of social presence, mediated social processes (i.e., interpersonal interaction, collaboration and community development) and learner support in the experiences of online learners.The results suggest a number of important findings with regard to social presence and learner support. First is an understanding of social presence as a quality of individual actors in the online environments. Social presence exists as a response to the limits of the technology and media in text-based online learning environments. Next, the study identifies social presence an essential component of online learning environments because it facilitates and supports mediated social processes including interpersonal interaction (or transaction), the development of relations between individuals and amongst groups of individuals, the progressive development of productive collaboration and the establishment and growth of community. Finally, the study highlights a revised model for learner support in text-based online learning environments which is informed by the development of learning communities and understanding of mediated social processes.