A Personal Agent Architecture for Task Automation in the Web of Data. Bringing intelligence to everyday tasks

Internet está evolucionando hacia la conocida como Live Web. En esta nueva etapa en la evolución de Internet, se pone al servicio de los usuarios multitud de streams de datos sociales. Gracias a estas fuentes de datos, los usuarios han pasado de navegar por páginas web estáticas a interacturar con aplicaciones que ofrecen contenido personalizado, basada en sus preferencias. Cada usuario interactúa a diario con multiples aplicaciones que ofrecen notificaciones y alertas, en este sentido cada usuario es una fuente de eventos, y a menudo los usuarios se sienten desbordados y no son capaces de procesar toda esa información a la carta. Para lidiar con esta sobresaturación, han aparecido múltiples herramientas que automatizan las tareas más habituales, desde gestores de bandeja de entrada, gestores de alertas en redes sociales, a complejos CRMs o smart-home hubs. La contrapartida es que aunque ofrecen una solución a problemas comunes, no pueden adaptarse a las necesidades de cada usuario ofreciendo una solucion personalizada. Los Servicios de Automatización de Tareas (TAS de sus siglas en inglés) entraron en escena a partir de 2012 para dar solución a esta liminación. Dada su semejanza, estos servicios también son considerados como un nuevo enfoque en la tecnología de mash-ups pero centra en el usuarios. Los usuarios de estas plataformas tienen la capacidad de interconectar servicios, sensores y otros aparatos con connexión a internet diseñando las automatizaciones que se ajustan a sus necesidades. La propuesta ha sido ámpliamante aceptada por los usuarios. Este hecho ha propiciado multitud de plataformas que ofrecen servicios TAS entren en escena. Al ser un nuevo campo de investigación, esta tesis presenta las principales características de los TAS, describe sus componentes, e identifica las dimensiones fundamentales que los defines y permiten su clasificación. En este trabajo se acuña el termino Servicio de Automatización de Tareas (TAS) dando una descripción formal para estos servicios y sus componentes (llamados canales), y proporciona una arquitectura de referencia. De igual forma, existe una falta de herramientas para describir servicios de automatización, y las reglas de automatización. A este respecto, esta tesis propone un modelo común que se concreta en la ontología EWE (Evented WEb Ontology). Este modelo permite com parar y equiparar canales y automatizaciones de distintos TASs, constituyendo un aporte considerable paraa la portabilidad de automatizaciones de usuarios entre plataformas. De igual manera, dado el carácter semántico del modelo, permite incluir en las automatizaciones elementos de fuentes externas sobre los que razonar, como es el caso de Linked Open Data. Utilizando este modelo, se ha generado un dataset de canales y automatizaciones, con los datos obtenidos de algunos de los TAS existentes en el mercado. Como último paso hacia el lograr un modelo común para describir TAS, se ha desarrollado un algoritmo para aprender ontologías de forma automática a partir de los datos del dataset. De esta forma, se favorece el descubrimiento de nuevos canales, y se reduce el coste de mantenimiento del modelo, el cual se actualiza de forma semi-automática. En conclusión, las principales contribuciones de esta tesis son: i) describir el estado del arte en automatización de tareas y acuñar el término Servicio de Automatización de Tareas, ii) desarrollar una ontología para el modelado de los componentes de TASs y automatizaciones, iii) poblar un dataset de datos de canales y automatizaciones, usado para desarrollar un algoritmo de aprendizaje automatico de ontologías, y iv) diseñar una arquitectura de agentes para la asistencia a usuarios en la creación de automatizaciones. ABSTRACT The new stage in the evolution of the Web (the Live Web or Evented Web) puts lots of social data-streams at the service of users, who no longer browse static web pages but interact with applications that present them contextual and relevant experiences. Given that each user is a potential source of events, a typical user often gets overwhelmed. To deal with that huge amount of data, multiple automation tools have emerged, covering from simple social media managers or notification aggregators to complex CRMs or smart-home Hub/Apps. As a downside, they cannot tailor to the needs of every single user. As a natural response to this downside, Task Automation Services broke in the Internet. They may be seen as a new model of mash-up technology for combining social streams, services and connected devices from an end-user perspective: end-users are empowered to connect those stream however they want, designing the automations they need. The numbers of those platforms that appeared early on shot up, and as a consequence the amount of platforms following this approach is growing fast. Being a novel field, this thesis aims to shed light on it, presenting and exemplifying the main characteristics of Task Automation Services, describing their components, and identifying several dimensions to classify them. This thesis coins the term Task Automation Services (TAS) by providing a formal definition of them, their components (called channels), as well a TAS reference architecture. There is also a lack of tools for describing automation services and automations rules. In this regard, this thesis proposes a theoretical common model of TAS and formalizes it as the EWE ontology This model enables to compare channels and automations from different TASs, which has a high impact in interoperability; and enhances automations providing a mechanism to reason over external sources such as Linked Open Data. Based on this model, a dataset of components of TAS was built, harvesting data from the web sites of actual TASs. Going a step further towards this common model, an algorithm for categorizing them was designed, enabling their discovery across different TAS. Thus, the main contributions of the thesis are: i) surveying the state of the art on task automation and coining the term Task Automation Service; ii) providing a semantic common model for describing TAS components and automations; iii) populating a categorized dataset of TAS components, used to learn ontologies of particular domains from the TAS perspective; and iv) designing an agent architecture for assisting users in setting up automations, that is aware of their context and acts in consequence.

Non contact inspection of the fatigue damage state of carbon fiber reinforced polymer by optical surface roughness measurements

This work presents the evaluation of a new non-contact technique to assess the fatigue damage state of CFRP structures by measuring surface roughness parameters. Surface roughness and stiffness degradation have been measured in CFRP coupons cycled with constant amplitude loads, and a Pearson?s correlation of 0.79 was obtained between both variables. Results suggest that changes on the surface roughness measured in strategic zones of components made of the evaluated CFRP, could be indicative of the level of damage due to fatigue loads. This methodology could be useful for other FRP due to similarities in the fatigue damage process.

A neuronal model of a global workspace in effortful cognitive tasks

A minimal hypothesis is proposed concerning the brain processes underlying effortful tasks. It distinguishes two main computational spaces: a unique global workspace composed of distributed and heavily interconnected neurons with long-range axons, and a set of specialized and modular perceptual, motor, memory, evaluative, and attentional processors. Workspace neurons are mobilized in effortful tasks for which the specialized processors do not suffice. They selectively mobilize or suppress, through descending connections, the contribution of specific processor neurons. In the course of task performance, workspace neurons become spontaneously coactivated, forming discrete though variable spatio-temporal patterns subject to modulation by vigilance signals and to selection by reward signals. A computer simulation of the Stroop task shows workspace activation to increase during acquisition of a novel task, effortful execution, and after errors. We outline predictions for spatio-temporal activation patterns during brain imaging, particularly about the contribution of dorsolateral prefrontal cortex and anterior cingulate to the workspace.

Positron-emission tomography studies of cross-modality inhibition in selective attentional tasks: closing the "mind's eye".

It is a familiar experience that we tend to close our eyes or divert our gaze when concentrating attention on cognitively demanding tasks. We report on the brain activity correlates of directing attention away from potentially competing visual processing and toward processing in another sensory modality. Results are reported from a series of positron-emission tomography studies of the human brain engaged in somatosensory tasks, in both "eyes open" and "eyes closed" conditions. During these tasks, there was a significant decrease in the regional cerebral blood flow in the visual cortex, which occurred irrespective of whether subjects had to close their eyes or were instructed to keep their eyes open. These task-related deactivations of the association areas belonging to the nonrelevant sensory modality were interpreted as being due to decreased metabolic activity. Previous research has clearly demonstrated selective activation of cortical regions involved in attention-demanding modality-specific tasks; however, the other side of this story appears to be one of selective deactivation of unattended areas.

Interval bias in 2AFC detection tasks: sorting out the artifacts

Proportion correct in two-alternative forcedchoice (2AFC) detection tasks often varies when the stimulus is presented in the first or in the second interval.Reanalysis of published data reveals that these order effects (or interval bias) are strong and prevalent, refuting the standard difference model of signal detection theory. Order effects are commonly regarded as evidence that observers use an off-center criterion under the difference model with bias. We consider an alternative difference model with indecision whereby observers are occasionally undecided and guess with some bias toward one of the response options. Whether or not the data show order effects, the two models fit 2AFC data indistinguishably, but they yield meaningfully different estimates of sensory parameters. Under indeterminacy as to which model governs 2AFC performance, parameter estimates are suspect and potentially misleading. The indeterminacy can be circumvented by modifying the response format so that observers can express indecision when needed. Reanalysis of published data collected in this way lends support to the indecision model. We illustrate alternative approaches to fitting psychometric functions under the indecision model and discuss designs for 2AFC experiments that improve the accuracy of parameter estimates, whether or not order effects are apparent in the data.

Electroencephalogram based Causality Graph Analysis in Behavior Tasks of Parkinson’s Disease Patients

Electroencephalographic (EEG) signals of the human brains represent electrical activities for a number of channels recorded over a the scalp. The main purpose of this thesis is to investigate the interactions and causality of different parts of a brain using EEG signals recorded during a performance subjects of verbal fluency tasks. Subjects who have Parkinson's Disease (PD) have difficulties with mental tasks, such as switching between one behavior task and another. The behavior tasks include phonemic fluency, semantic fluency, category semantic fluency and reading fluency. This method uses verbal generation skills, activating different Broca's areas of the Brodmann's areas (BA44 and BA45). Advanced signal processing techniques are used in order to determine the activated frequency bands in the granger causality for verbal fluency tasks. The graph learning technique for channel strength is used to characterize the complex graph of Granger causality. Also, the support vector machine (SVM) method is used for training a classifier between two subjects with PD and two healthy controls. Neural data from the study was recorded at the Colorado Neurological Institute (CNI). The study reveals significant difference between PD subjects and healthy controls in terms of brain connectivities in the Broca's Area BA44 and BA45 corresponding to EEG electrodes. The results in this thesis also demonstrate the possibility to classify based on the flow of information and causality in the brain of verbal fluency tasks. These methods have the potential to be applied in the future to identify pathological information flow and causality of neurological diseases.

Evaluating EmotiBlog robustness for sentiment analysis tasks

EmotiBlog is a corpus labelled with the homonymous annotation schema designed for detecting subjectivity in the new textual genres. Preliminary research demonstrated its relevance as a Machine Learning resource to detect opinionated data. In this paper we compare EmotiBlog with the JRC corpus in order to check the EmotiBlog robustness of annotation. For this research we concentrate on its coarse-grained labels. We carry out a deep ML experimentation also with the inclusion of lexical resources. The results obtained show a similarity with the ones obtained with the JRC demonstrating the EmotiBlog validity as a resource for the SA task.

Using WordNet relations and semantic classes in information retrieval tasks

In this paper we explore the use of semantic classes in an existing information retrieval system in order to improve its results. Thus, we use two different ontologies of semantic classes (WordNet domain and Basic Level Concepts) in order to re-rank the retrieved documents and obtain better recall and precision. Finally, we implement a new method for weighting the expanded terms taking into account the weights of the original query terms and their relations in WordNet with respect to the new ones (which have demonstrated to improve the results). The evaluation of these approaches was carried out in the CLEF Robust-WSD Task, obtaining an improvement of 1.8% in GMAP for the semantic classes approach and 10% in MAP employing the WordNet term weighting approach.

A combined probabilistic framework for learning gestures and actions

In this paper we introduce a probabilistic approach to support visual supervision and gesture recognition. Task knowledge is both of geometric and visual nature and it is encoded in parametric eigenspaces. Learning processes for compute modal subspaces (eigenspaces) are the core of tracking and recognition of gestures and tasks. We describe the overall architecture of the system and detail learning processes and gesture design. Finally we show experimental results of tracking and recognition in block-world like assembling tasks and in general human gestures.

Human-computer interaction for optical music recognition tasks

The need to digitise music scores has led to the development of Optical Music Recognition (OMR) tools. Unfortunately, the performance of these systems is still far from providing acceptable results. This situation forces the user to be involved in the process due to the need of correcting the mistakes made during recognition. However, this correction is performed over the output of the system, so these interventions are not exploited to improve the performance of the recognition. This work sets the scenario in which human and machine interact to accurately complete the OMR task with the least possible effort for the user.

Calibración de cámaras de tiempo de vuelo: ajuste adaptativo del tiempo de integración y análisis de la frecuencia de modulación

La percepción de profundidad se hace imprescindible en muchas tareas de manipulación, control visual y navegación de robots. Las cámaras de tiempo de vuelo (ToF: Time of Flight) generan imágenes de rango que proporcionan medidas de profundidad en tiempo real. No obstante, el parámetro distancia que calculan estas cámaras es fuertemente dependiente del tiempo de integración que se configura en el sensor y de la frecuencia de modulación empleada por el sistema de iluminación que integran. En este artículo, se presenta una metodología para el ajuste adaptativo del tiempo de integración y un análisis experimental del comportamiento de una cámara ToF cuando se modifica la frecuencia de modulación. Este método ha sido probado con éxito en algoritmos de control visual con arquitectura ‘eye-in-hand’ donde el sistema sensorial está compuesto por una cámara ToF. Además, la misma metodología puede ser aplicada en otros escenarios de trabajo.

Monitoring an earthfill dam using differential SAR interferometry: La Pedrera dam, Alicante, Spain

Surface displacement at the dykes of La Pedrera reservoir (SE Spain) has been measured by satellite differential Synthetic Aperture Radar (SAR) interferometry. At the main dyke, a displacement of about 13 cm along the satellite line of sight has been estimated between August 1995 and May 2010, from a dataset composed by ERS-1, ERS-2 and Envisat-ASAR images. Two independent short-term processing tasks were also carried out with ERS-2/Envisat-ASAR (from June 2008 to May 2010) and TerraSAR-X (from August 2008 to June 2010) images which have shown similar spatial and temporal displacement patterns. The joint analysis of historical instrument surveys and DInSAR-derived data has allowed the identification of a long-term deformation process which is reflected at the dam's surface and is also clearly recognizable in the inspection gallery. The plausible causes of the displacements measured by DInSAR are also discussed in the paper. Finally, DInSAR data have been used to compute the long-term settlement of La Pedrera dam, showing a good agreement with external studies. Consequently, this work demonstrates the integration of DInSAR with in-situ techniques which helps provide a complete spatial vision of the displacements in the dam thereby helping to differentiate the causal mechanisms.

Active Visual Features Based on Events to Guide Robot Manipulators in Tracking Tasks

Traditional visual servoing systems do not deal with the topic of moving objects tracking. When these systems are employed to track a moving object, depending on the object velocity, visual features can go out of the image, causing the fail of the tracking task. This occurs specially when the object and the robot are both stopped and then the object starts the movement. In this work, we have employed a retina camera based on Address Event Representation (AER) in order to use events as input in the visual servoing system. The events launched by the camera indicate a pixel movement. Event visual information is processed only at the moment it occurs, reducing the response time of visual servoing systems when they are used to track moving objects.