Tesis para optar al grado de Doctora en Ciencias, Mención Computación / Buscar contenido multimedia es una de las tareas más comunes que los usuarios realizan en la Web. Actualmente, los motores de búsqueda en la Web han mejorado la precisión de sus búsquedas de contenido multimedia y ahora brindan una mejor experiencia de usuarios. Sin embargo, estos motores aún no logran obtener resultados precisos para consultas que no son comunes, y consultas que se refieren a conceptos abstractos. En ambos escenarios, la razón principal es la falta de información preliminar. Esta tesis se enfoca en mejorar la recuperación de información multimedia en la Web usando datos generados a partir de la interacción entre usuarios y recursos multimedia. Para eso, se propone mejorar la recuperación de información multimedia desde dos perspectivas: (1) extrayendo conceptos relevantes a los recursos multimedia, y (2) mejorando las descripciones multimedia con datos generados por el usuario. En ambos casos, proponemos sistemas que funcionan independientemente del tipo de multimedia, y del idioma de los datos de entrada. En cuanto a la identificación de conceptos relacionados a objetos multimedia, desarrollamos un sistema que va desde los resultados de búsqueda específicos de la consulta hasta los conceptos detectados para dicha consulta. Nuestro enfoque demuestra que podemos aprovechar la vista parcial de una gran colección de documentos multimedia para detectar conceptos relevantes para una consulta determinada. Además, diseñamos una evaluación basada en usuarios que demuestra que nuestro algoritmo de detección de conceptos es más sólido que otros enfoques similares basados en detección de comunidades. Para mejorar la descripción multimedia, desarrollamos un sistema que combina contenido audio-visual de documentos multimedia con información de su contexto para mejorar y generar nuevas anotaciones para los documentos multimedia. Específicamente, extraemos datos de clicks de los registros de consultas y usamos las consultas como sustitutos para las anotaciones manuales. Tras una primera inspección, demostramos que las consultas proporcionan una descripción concisa de los documentos multimedia. El objetivo principal de esta tesis es demostrar la relevancia del contexto asociado a documentos multimedia para mejorar el proceso de recuperación de documentos multimedia en la Web. Además, mostramos que los grafos proporcionan una forma natural de modelar problemas multimedia. / Fondef D09I-1185, CONICYT-PCHA/Doctorado Nacional/2013-63130260, Apoyo a estadías corta de la Escuela de Postgrado de la U. de Chile, y el Núcleo Milenio CIWS

Im akademischen Bereich sind in Social-Software-Anwendungen wie Connotea, CiteULike und BibSonony umfangreiche Sammlungen von nutzergenerierten Metadaten entstanden. Im Vergleich zu kontrollierten Vokabularen, wie der Schlagwortnormdatei, handelt es sich dabei um personalisierte und in weiten Teilen „chaotische“ Inhaltserschließung. An der Universitätsbibliothek Mannheim wurde in einem DFG-Projekt untersucht, inwieweit das Potential dieser Art von Metadaten für eine bessere und nutzerorientierte Präsentation von Informationsressourcen eingesetzt werden kann. Ein Kernstück der Untersuchung war die Analyse von Tag-Daten des Systems BibSonomy. Es zeigte sich, dass nicht nur die mangelnde semantische Strukturiertheit der Tags, sondern auch ihre heterogene Gestalt einen limitierenden Faktor für die Verwendung in der bibliothekarischen Sacherschließung darstellt. Der Beitrag gibt anhand von Beispielen Einblick in das qualitative und strukturelle Chaos der untersuchten Tags und fasst die Ergebnisse des Projekts zusammen.

Social Tagging und soziale Netzwerke sind zentrale Bausteine des Web 2.0 und Enterprise 2.0. In diesem Beitrag werden die sozialen Aspekte von Social Tagging beleuchtet und ein Ansatz aufgeführt, um in Folksonomies Personen mit ähnlichen Interessen zu finden. Ferner wird ein Tagging-Framework beschrieben, das im Use Case Alexandria im Rahmen des BMWi-Projekts Theseus entstanden ist.

Acoustic telemetry is an important tool for studying the movement patterns, behaviour, and site fidelity of marine organisms; however, its application is challenged in coral reef environments where complex topography and intense environmental noise interferes with acoustic signals, and there has been less study. Therefore, it is particularly critical in coral reef telemetry studies to first conduct a long-term range test, a tool that provides informa- tion on the variability and periodicity of the transmitter detection range and the detection probability. A one-month range test of a coded telemetric system was conducted prior to a large-scale tagging project investigating the movement of approximately 400 fishes from 30 species on offshore coral reefs in the central Red Sea. During this range test we determined the effect of the following factors on transmitter detection efficiency: distance from receiver, time of day, depth, wind, current, moon-phase and temperature. The experiment showed that biological noise is likely to be responsible for a diel pattern of -on average- twice as many detections during the day as during the night. Biological noise appears to be the most important noise source in coral reefs overwhelming the effect of wind-driven noise, which is important in other studies. Detection probability is also heavily influenced by the location of the acoustic sensor within the reef structure. Understanding the effect of environmental factors on transmitter detection probability allowed us to design a more effective receiver array for the large-scale tagging study.

The present study measured the accuracy of automated grammatical tagging software as compared to manual tagging in Spanish-speaking children's personal and fictional event narrative language samples. Studies have identified articles, clitic (contracted with a verb) pronouns, and verbs as clinical markers for language impairment in Spanish-speaking children. Automated grammatical tagging software may aid in the rapid identification of these grammatical markers. Grammatical morphemes of 30 first and fourth grade children's personal and fictional event narrative samples were tagged and compared with their respective manually tagged samples. The accuracy of word-level coding averaged 91%, and similar accuracy was found for clinically significant tags. Automated grammatical analysis has the potential to accurately identify clinically relevant grammatical forms in samples from children who speak Spanish.

There are currently different efforts to use Supervised Neural Networks (NN) to automatically label damages on images of above ground infrastructure (buildings made of concrete) taken after an earthquake. The goal of the supervised NN is to classify raw input data according to the patterns learned from an input training set. This input training data set is usually supplied by experts in the field, and in the case of this project, structural engineers carefully and mostly manually label these images for different types of damage. The level of expertise of the professionals labeling the training set varies widely, and some data sets contain pictures that different people have labeled in different ways when in reality the label should have been the same. Therefore, we need to get several experts to evaluate the same data set; the bigger the ground truth/training set the more accurate the NN classifier will be. To evaluate these variations among experts, which can be considered equal to the task of evaluating the quality of the expert, using probabilistic theory we first need to implement a tool able to compare different images classified by different experts and apply a certainty level to the experts tagged labels. This master's thesis implements this comparative tool. We also decided to implement the comparative tool using parallel programming paradigms since we foresee that it will be used to train multiple young engineering students/professionals or even novice citizen volunteers (“trainees”) during after-earthquake meetings and workshops. The implementation of this software tool involves selecting around 200 photographs tagged by an expert with proven accuracy (“ground truth”) and comparing them to files tagged by the trainees. The trainees are then provided with instantaneous feedback on the accuracy of their damage assessment. The aforementioned problem of evaluating trainee results against the expert is not as simple as comparing and finding differences between two sets of image files. We anticipate challenges in that each trainee will select a slightly different sized area for the same occurrence of damage, and some damage-structure pairs are more difficult to recognize and tag. Results show that we can compare 500 files in 1.5 seconds which is an improvement of 2x faster compared to sequential implementation.

Yes / The domain of cultural variations in interpersonal communication is becoming increasingly important in various areas, including human-human interaction (e.g. business settings) and humancomputer interaction (e.g. during simulations, or with social robots). User generated content (UGC) in social media can provide an invaluable source of culturally diverse viewpoints for supporting the understanding of cultural variations. However, discovering and organizing UGC is notoriously challenging and laborious for humans, especially in ill-defined domains such as culture. This calls for computational approaches to automate the UGC sensemaking process by using tagging, linking and exploring. Semantic technologies allow automated structuring and qualitative analysis of UGC, but are dependent on the availability of an ontology representing the main concepts in a specific domain. For the domain of cultural variations in interpersonal communication, no ontological model exists. This paper presents the first such ontological model, called AMOn+, which defines cultural variations and enables tagging culture-related mentions in textual content. AMOn+ is designed based on a novel interdisciplinary approach that combines theoretical models of culture with crowdsourced knowledge (DBpedia). An evaluation of AMOn+ demonstrated its fitness-for-purpose regarding domain coverage for annotating culture-related concepts mentioned in text corpora. This ontology can underpin computational models for making sense of UGC.

No description available.

Given the ecological importance of face recognition, it is not surprising that the visual system is capable of processing faces with remarkable efficiency. When presented with a face, information is rapidly extracted to detect and categorize it as a face, followed by face-specific information such as age, gender, and identity. According to cognitive and neural models, the processes underlying face recognition encompass a sequence of steps that begin with a perceptual or visual analysis followed by more image-invariant and identity-selective representations. Importantly, it is only familiar faces for which we have acquired long-term face memories that reach the final stages of identity processing to permit robust, image-invariant behavioural recognition. A key aspect of face processing is that it is fast and automatic. This can be said for both high-level categorization (i.e., detecting that a stimulus is a face) and for encoding at the identity-level. The purpose of these experiments was to use novel electrophysiological and psychophysical techniques to characterize these fast and automatic categorization processes. Experiment 1 and 2 used an implicit visual discrimination paradigm (fast periodic visual stimulation; FPVS) combined with electroencephalography (EEG) to isolate identity-specific neural responses to a personally familiar face, the own-face, and an unfamiliar stranger face. Experiment 1 showed that identity-specific responses recorded over the occipito-temporal region were stronger for a personally familiar face compared to the unfamiliar control identity, while the response to the own-face was even greater than to a personally familiar friend. In Experiment 2, identity-specific responses for a given identity were measured in participants both before and after real-world familiarization. As expected, the results showed a significant increase in the identity-specific response once participants became personally familiar with the test identities. In Experiment 3, we used saccadic eye movements to estimate the lower bounds of the speed of face categorization, and in particular to investigate the question of whether this categorization occurs during early feedforward processing. The results support the view that information needed to detect and selectively respond to face stimuli happens during the earliest visual processing. Collectively, these studies provide additional insight on the mechanisms underlying rapid and automatic face detection and face identity recognition. / Graduate

Yes / Social media is a platform to express one’s view in real time. This real time nature of social media makes it an attractive tool for disaster management, as both victims and officials can put their problems and solutions at the same place in real time. We investigate the Twitter post in a flood related disaster and propose an algorithm to identify victims asking for help. The developed system takes tweets as inputs and categorizes them into high or low priority tweets. User location of high priority tweets with no location information is predicted based on historical locations of the users using the Markov model. The system is working well, with its classification accuracy of 81%, and location prediction accuracy of 87%. The present system can be extended for use in other natural disaster situations, such as earthquake, tsunami, etc., as well as man-made disasters such as riots, terrorist attacks etc. The present system is first of its kind, aimed at helping victims during disasters based on their tweets.