Acoustic source separation is a relatively recent topic of signal processing which aims to simultaneously separate many acoustic sources recorded through one or more microphones. Such a problem was formulated to emulate the natural capability of the human auditory system which is able to recognize and enhance the sound coming from a particular source. Addressing this problem is of high interest in the automatic speech recognition (ASR) community since it would improve the effectiveness of a natural human-machine interaction. Among numerous methods of multichannel blind source separation techniques, those based on the Independent Component Analysis (ICA) applied in the frequency-domain [81] are the most investigated, due to their straightforward physical interpretation and computational efficiency. In spite of recent developments many issues still need to be address to make such techniques robust in adverse conditions, such as high reverberation, ill-conditioning and occurrence of permutations. Furthermore, most of the proposed BSS methods are computationally expensive and not feasible for a real-time implementation. This PhD thesis describes a research activity in the robust separation of acoustic sources in adverse environment. A new framework of blind and semi-blind techniques is proposed which allows source localization and separation even in highly reverberant environment and with realtime constraint. For each proposed technique, theoretical and practical issues are discussed and a comparison with alternative state-of-art methods is provided. Furthermore, the robustness of the proposed framework is validated implementing two real-time blind and semi-blind systems which are tested in challenging real-world scenarios.

People belong to different communities: business communities, Web 2.0 communities, religious communities, scientific communities, just to name a few. Everyone can belong to and acquire experience in more than one community. This experience is related to the community activity and comes in the form of best practices, behavior, implicit (tacit) knowledge, ways of using artifacts, etc. All these accumulates and evolves over time and slowly becomes a part of the culture of the community. If community activity is very specific, it can be reflected also in the specificity of the community culture. Newcomers in such community might suffer from what is called “culture shock†, i.e. a feeling of confusion when not able to grasp what is common for old-timers. This occurs because part of the community culture is not explicit, i.e. not readily available, and it is very hard to extract valuable information from it. Such information can be used for increasing economic and social benefits of the community members (e.g., for performing recurring tasks, easier integration of newcomers, better quality of life). Moreover, the awareness of the community culture could help the community to handle the turnover of members and structural changes, while preserving the culture. All these introduce the need for transfer of culture between or within communities. Currently, there is no domain-independent approach for discovering, representing, transferring, and preserving community culture. Moreover, taking into account the amount of information accumulated by communities, computer aided tools for such representation and transfer are of utmost importance. A key property of such tools should be their non-intrusiveness, i.e. they must be as much integrated in the community practices as possible. Research challenges in solving these problems include, but are not limited to: 1) providing a generic approach for dealing with community culture; 2) designing a framework and computer aided supporting tools for transferring culture; 3) implementing the framework, applying and evaluating it in different domains. This thesis addresses the problem of culture transfer. First, we formalize the notion of culture, which includes behavior, knowledge, artifacts, best practices, etc., and provide a classification of problems that involve culture. Second, using this formalism, we propose the Implicit Culture Framework, which is an agent-based framework for transferring behavior among community members or among communities. Then we describe three applications developed using the IC-Service in the domain of recommendation systems: a system for web search, a system for software pattern selection, and a system for web service discovery. Finally, we present the results of the evaluation of the applications with real users and with ad-hoc user models.

Security has been a growing concern for large organizations, especially financial and gov- ernmental institutions, as security breaches in the systems they depend have repeatedly resulted in losses of billions per year, and this cost is on the rise. A primary reason for these breaches is the “socio-technical” nature of today’s systems that consist of an amal- gam of social and human actors, processes, technology and infrastructure. We refer to such systems as Socio-Technical Systems (STSs). Finding secure solutions for STSs is a difficult and error-prone task because of their heterogeneity and complexity. The thesis proposes a holistic security requirements analysis framework which catego- rizes system security concerns into three layers, including a social layer (social actors and business processes), a software layer (software applications that support the social layer) and an infrastructure layer (physical infrastructure, hardware, and devices). Within each layer, security requirements are elicited, and security mechanisms are designed to satisfy the security requirements. In particular, a cross-layer support link is defined to capture how security mechanisms deployed at one layer influence security requirements of the next layer down, allowing us to systematically and iteratively analyze security for all three layers and eventually produce holistic security solutions for the systems. To ensure the quality of the analysis of our approach and to promote practical adoption of the three-layer approach, the thesis includes two additional components. Firstly, we propose a holistic attack analysis, which takes an attacker’s perspective to explore realistic attacks that can happen to a system and thus contributes to the identification of critical security requirements. This approach consists of an attack strategy identification method which analyzes attacker’s alternative malicious intentions, and an attack strategy operationalization method which analyzes realistic attack actions that can be performed by attackers. Secondly, the thesis proposes a systematic approach for selecting and applying security patterns, which describe proven security solutions to known security problems. As such, analysts with little security knowledge can efficiently leverage reusable security knowledge to operationalize security requirements in terms of security mechanisms. This approach also allows us to systematically analyze and enforce the impact of deployed security mechanisms on system functional specifications. We have developed a prototype tool, which implements the formalized analysis methods of our three-layer framework and enables the semi-automatic application of our proposal. With the help of the tool, we apply our framework to two large-scale case studies so as to validate the efficacy of our approach.

This thesis positions itself within the stream of research on HCI for sport and addresses the topic of designing wearable devices for sport. To date, the design of wearables for sport has focused on the measurable aspects of performance such as speed, heartbeat and calories burnt. Such design is driven by the possibilities offered by the miniaturisation of components and the trend to have a healthy lifestyle. The conjunction of these two trends, has created a breeding ground for technologies that offer self-tracking to improve personal fitness, health and wellbeing. Although these kinds of devices have great success on the market, several studies have shown poor long-term adoption, with people generally ceasing to use their devices around six months from the time of purchase. This thesis argues that the wearables produced until now do not address the full range of needs that sportspeople have and so aims to design wearables on the basis of a thorough understanding of the sport practice. The leading research question in this work was: what are the elements to consider for the design of useful, acceptable and desirable wearable devices for sport? This broad research question was then operationalised in two sub-questions: what elements constitute the sport practice?; and how can wearable devices support such practice? By adopting a practice perspective and a subsequent research methodology based on situatedness, embodiment, and co-design, it was possible to identify aspects of sport other than performance. Emotions, trust and community values emerged as pivotal aspects of the climbing experience. These findings led to the design of wearables for augmenting the interpersonal communication of the actors involved. This introduces a new role for wearables supporting sportspeople, which as a facilitator of expertise rather than a tracker of activity. The main contribution of this thesis is the articulation of a conceptual framework for the design of wearables for outdoor sports, with the goal of better acceptance and long-term adoption. The conceptual framework outlined here breaks down the complexity of the sport practice by identifying the elements that define it (i.e. type of performance, emotional involvement, social dynamics, physical context, values) and articulating their orchestration with product design aspects (such as ergonomics, comfort, and perceptibility) and the cultural value of wearing an artefact on the body.

Automatic behavior analysis lies at the intersection of different social and technical research domains. The interdisciplinarity of the field, provides researchers with the means to study the manifestations of human constructs, such as personality. A branch of human behavior analyis, the study of personality provides insight into the cognitive and psychological construction of the human being. Research in personality psychology, advances in com- puting power and the development of algorithms, have made it possible to analyze existing data in order to understand how people express their own personality, perceive others’, and what are the variables that influence its manifestation. We are pursuing this line of research because insights into the personality of the user can have an impact on how we interact with technology. Incorporating research on personality recogniton, both from a cognitive as well as an engineering perspective, into computers could facilitate the interactions between humans and machines. Previous attempts on personality recognition have focused on a variety of different corporas (ranging from text to audiovisual data), different scenarios (interviews, meetings), different channels of communication (audio, video, text) and different subsets of personality traits (out of the five ones present in the Big Five Model: Extraversion, Agreeableness, Conscientiousness, Emotional Stability and Creativity). Our work builds on previous research, by considering simple acoustic and visual non-verbal features extracted from multimodal data, but doesn’t fail to bring novelties: we consider previously uninvestigated scenarios, and at the same time, all of the five personality traits and not just a subset.

The employment of visual sensor networks in surveillance systems has brought in as many challenges as disadvantages. While the integration of multiple cameras into a network has the potential advantage of fusing complementary observations from sensors and enlarging visual coverage, it also increases the complexity of tracking tasks and poses challenges to system scalability. The research work in this thesis addresses the problem of building an efficient and scalable multi-camera tracking system that (i) adapts, in real time, to the dynamics of the monitored scenario and (ii) attempts to maximize usage and sharing of available sensing and processing resources. To perform reliable tracking, a preliminary step is fast and accurate people detection, for which we propose a locus-based probabilistic occupancy map (LPOM). The LPOM computes the probability of targets being in the map by using only motion information plus calibration data. To make the tracking system more scalable, we present a decentralized multi-camera multi-people tracking framework with a three-layer architecture, in which we formulate the overall task (i.e. tracking all people using all available cameras) as a vision based state estimation problem and aim to maximize utility and sharing of available sensing and processing resources. By exploiting the geometric relations between sensing geometry and people's positions, our method is able to dynamically and adaptively partition the overall task into a number of nearly independent subtasks with the aid of information theory, each of which tracks a subset of people with a subset of cameras (or agencies). The method hereby reduces task complexity dramatically and helps to boost parallelization and maximize the system's real time throughput and reliability while accounting for intrinsic uncertainty induced, e.g., by visual clutter, occlusion, and illumination changes. Moreover, we propose a preliminary information theoretical framework, in which we apply task-driven polling strategies that control the sensing process to minimize the number of observations to be processed while maximizing their expected impact on estimation process, and use parameter adaptation techniques to reallocate computational resources dynamically and opportunistically according to task complexity and measured evidence. For each proposed approach we carry out a number of experiments and demonstrate its efficiency and advantages.

Sociotechnical systems are systems of systems where social, technical, and organizational systems interact with each other to satisfy their requirements. The interplay of social and technical systems blurs the borders in between them, and the constant change within and outside the sociotechncial systems create difficulties to manage the overall evolution. This thesis explores the methods to model, analyse, and evolve the requirements of sociotechnical systems. We propose a systematic design process and a formal language to aid social systems refine their requirements into not other requirements but also social interactions to generate system as well as interaction specifications. Although such specifications are useful to generate interaction protocols among systems, they haven’t been investigated in detail by the requirements engineering community. We then explore the design space created during the design process with artificial intelligence planing to discover sequence of actions to satisfy requirements with minimal cost. We adopt an iterative approach for handling requirements evolution and focus on the problem of selecting the optimal set of requirements for the next release. We capture synergies among requirements in goal-oriented requirements models and transform the next release problem into a multi-objective satisfiability modulo theories/optimization modulo theories problem and solve it using an external reasoner. We apply a similar approach for risk analysis using goal models. We model goals, risks, and treatments in three layers and solve multi-objective risk analysis problem with SMT/OMT reasoning. We evaluate our proposal with self-evaluation studies, a case study and scalability experiments and report results. The novelty of these two approaches is the combination of satisfiability analysis with multi-objective optimization for goal models.

The Linked Open Data (LOD) cloud is currently a primary source of background knowledge for tasks in a wide variety of domains and across many scientific fields. The structured nature and the usage of well-defined open standards make it convenient to contribute to and build upon. However, since the major part of the LOD is ultimately crowdsourced and mostly populated and updated manually, some of the content in the LOD can become stale, inconsistent and lack coverage. Social media, on the other hand, uniquely allow the real world events to be accurately reflected with little or no delay in the form of posts and profile updates. A major downside of this vibrant source of knowledge that is contained in the social media is its lack of structure, significant noisiness and restrictive APIs that make it hard to extract, analyze and use it in the downstream tasks. In this thesis, I present the task of linking entities in a knowledge base (KB) to the corresponding social media profiles as an attempt to bridge the structured LOD cloud and the vibrant social media. As will be shown, such linking allows knowledge transfer between the two worlds: on the one hand, enabling the Semantic Web practitioners to harvest this vast amount of valuable, up-to-date data from the social media; on the other hand, the social media researchers can use the structured LOD knowledge much more efficiently, simplifying the pipelines and improving performance for tasks such as Type Prediction, Entity Linking, and User Profiling. I implement such knowledge transfer using DBpedia as a KB, since it is a cornerstone dataset in the LOD, and Twitter as a social media, due to its popularity and relative accessibility. However, approaches developed here are designed to be general and could be applied to other social media and KBs. To this end, firstly, I introduce SocialLink - a project designed to link KBs to social media profiles. SocialLink consists of (i) a linking approach that is able to produce high-quality entity-profile pairs, (ii) a LOD-compliant dataset of alignments between DBpedia and Twitter, (iii) the Social Media Toolkit system providing additional functionality on top of SocialLink. SocialLink employs a custom deep neural network-based architecture designed to efficiently exploit many modalities of data representing entities and profiles within DBpedia and Twitter. In second, I demonstrate how SocialLink can facilitate tasks in both Semantic Web and Social Media Analysis. In particular, I employ the abovementioned knowledge transfer to achieve state-of-the-art performance in Type Prediction task on DBpedia. Additionally, SocialLink is used to infer user interests on Twitter and to implement a novel approach that I proposed to prevent such inference. Finally, the Entity Linking capabilities of SocialLink are exploited to augment the social media management application called Pokedem and to provide an additional performance boost to a conventional Entity Linking pipeline achieving the second-best performance in EVALITA 2016 competition.

In this thesis we investigate the automatic analysis of agreement and disagreement in political documents. Our focus is on the comparison of statements about specific topics extracted from documents with no direct interaction (e.g. electoral speeches or political manifestos), in which politicians may express, sometime in an implicit way, their position. This is a challenging task, made difficult also due to the lack of annotated resources. Our contribution can be divided into two main areas. The first one is the creation of manually and automatically annotated corpora for the task (pairs of statements annotated for agreement or disagreement from different sources). The second one is a Natural Language Processing (NLP) pipeline for the automatic (supervised) classification of agreement and disagreement. This pipeline involves a novel approach to extract well-defined and accurate topics based on key-concept clusters, and two classifiers to identify the pairs of statements in agreement and disagreement (or holding no relation) according to a wide set of features, such as sentiment, entailment, and semantic representation of the topics. We think that our findings can effectively support political science researchers dealing with an increasing amount of digital data, providing insight into similarities and differences in ideologies.

Enforcing legal compliance into software systems is a non-trivial task that requires an interdisciplinary approach. This thesis presents a new methodology for legal compliance checking against European legal provisions, namely the EU Data Protection Directive, the EU General Data Protection Regulation and the revised EU Payment Services Directive. We propose two types of compliance checking mechanisms that should be exploited at design-time or run-time. The former is based on security policy analysis of access control policies. The later is built on top of an approach to synthesizing run-time monitors for workflow-driven applications. Our contributions include a comprehensive methodology for legal compliance checking, the formalization of the regulations and the prototype tool of the implemented compliance methodology.