The Datagram Transport Layer Security (DTLS) protocol is highly vulnerable to a form of denial-of-service attack (DoS), aimed at establishing a high number of invalid, half-open, secure sessions. Moreover, even when the efficient pre-shared key provisioning mode is considered, the key storage on the server side scales poorly with the number of clients. SICS Swedish ICT has designed a security architecture that efficiently addresses both issues without breaking the current standard.
This article presents Axiom, a DTLS-based approach to efficiently secure multicast group communication among IoT-constrained devices. Axiom provides an adaptation of the DTLS record layer, relies on key material commonly shared among the group members, and does not require one to perform any DTLS handshake. We made a proof-of-concept implementation of Axiom based on the tinyDTLS library for the Contiki OS and used it to experimentally evaluate performance of our approach on real IoT hardware. Results show that Axiom is affordable on resource-constrained platforms and performs significantly better than related alternative approaches.
ICT is becoming a fundamental and pervasive component of critical infrastructures (CIs). Despite the advantages that it brings about, ICT also exposes CIs to a number of security attacks that can severely compromise human safety, service availability and business interests. Although it is vital to ensure an adequate level of security, it is practically infeasible to counteract all possible attacks to the maximum extent. Thus, it is important to understand attacks' impact and rank attacks according to their severity. We propose SEA++, a tool for simulative evaluation of attack impact based on the INET framework and the OMNeT++ platform. Rather than actually executing attacks, SEA++ reproduces their effects and allows to quantitatively evaluate their impact. The user describes attacks through a high-level description language and simulates their effects without any modification to the simulation platform. We show SEA++ capabilities referring to different attacks carried out against a traffic light system.
Software Defined Networking (SDN) has been recently introduced as a new communication paradigm in computer networks. By separating the control plane from the data plane and entrusting packet forwarding to straightforward switches, SDN makes it possible to deploy and run networks which are more flexible to manage and easier to configure. This paper describes a set of extensions for the INET framework, which allow researchers and network designers to simulate SDN architectures and evaluate their performance and security at design time. Together with performance evaluation and design optimization of SDN networks, our extensions enable the simulation of SDN-based anomaly detection and mitigation techniques, as well as the quantitative evaluation of cyber-physical attacks and their impact on the network and application. This work is an ongoing research activity, and we plan to propose it for an official contribution to the INET framework.
As supervised machine learning methods for addressing tasks in natural language processing (NLP) prove increasingly viable, the focus of attention is naturally shifted towards the creation of training data. The manual annotation of corpora is a tedious and time consuming process. To obtain high-quality annotated data constitutes a bottleneck in machine learning for NLP today. Active learning is one way of easing the burden of annotation. This paper presents a first probe into the NLP research community concerning the nature of the annotation projects undertaken in general, and the use of active learning as annotation support in particular.
Electric roads have had a pleasant journey through the research landscape but are now about to enter the 'valley of death.' The basic technologies for dynamic power transfer from the road to vehicles in motion has been developed through various research projects across the globe, largely supported by public funding. Electric road systems (ERS) will soon be tested on public roads, but is still a long way from constituting a large-scale commercial system. While ERS has gained recognition as a technological solution, few studies address the necessary system transition from a holistic perspective. This article addresses this gap by presenting the state of the art of ERS and examining future use case scenarios and stakeholder implications. The purpose of this article is accordingly to examine how ERS not only constitutes a technical development challenge, but also radically increases technical, business, and systems complexity. This article illustrates how ERS will likely evolve from a system to a system of systems and the likely changes in the business and system architecture occurring during this transformation will be analyzed. Finally, future challenges will be discussed.
Code quality remains an abstract concept that fails to get traction at the business level. Consequently, software companies keep trading code quality for time-to-market and new features. The resulting technical debt is estimated to waste up to 42% of developers' time. At the same time, there is a global shortage of software developers, meaning that developer productivity is key to software businesses. Our overall mission is to make code quality a business concern, not just a technical aspect. Our first goal is to understand how code quality impacts 1) the number of reported defects, 2) the time to resolve issues, and 3) the predictability of resolving issues on time. We analyze 39 proprietary production codebases from a variety of domains using the CodeScene tool based on a combination of source code analysis, version-control mining, and issue information from Jira. By analyzing activity in 30,737 files, we find that low quality code contains 15 times more defects than high quality code. Furthermore, resolving issues in low quality code takes on average 124% more time in development. Finally, we report that issue reso-lutions in low quality code involve higher uncertainty manifested as 9 times longer maximum cycle times. This study provides evi-dence that code quality cannot be dismissed as a technical concern. With 15 times fewer defects, twice the development speed, and substantially more predictable issue resolution times, the business advantage of high quality code should be unmistakably clear.
Anonymization methods are one potential way of alleviating the risks of capturing personal information during data collections. The work presented here is based on one such method that, in turn, is based on generating images through machine learning to replace the original images. The chosen method merges both the original image and the generated one resulting in a risk of information from the original image leaking through to the final result. Here a possible approach to measure how much influence the original image has on the final product is presented .
Thermal contact resistances between a silver metallized SiC chip and a Molybdenum substrate and between the Molybdenum substrate and bulk Copper were measured in a heat transfer experiment. An experimental method to separate thermal contact resistances in a multilayer heat transfer path was used to extract the layer-specific contact resistances. The experimental results were compared with theory based calculations and also with 3-D computational fluid dynamics (CFD) simulation results. The results show significant pressure dependence of the thermal contact resistance and the results show higher thermal contact resistance per unit area between the bulk SiC chip and Molybdenum than between Molybdenum and bulk Copper.
Use of the serverless paradigm in cloud application development is growing rapidly, primarily driven by its promise to free developers from the responsibility of provisioning, operating, and scaling the underlying infrastructure. However, modern cloud-edge infrastructures are characterized by large numbers of disparate providers, constrained resource devices, platform heterogeneity, infrastructural dynamicity, and the need to orchestrate geographically distributed nodes and devices over public networks. This presents significant management complexity that must be addressed if serverless technologies are to be used in production systems. This position paper introduces COGNIT, a major new European initiative aiming to integrate AI technology into cloud-edge management systems to create a Cognitive Cloud reference framework and associated tools for serverless computing at the edge. COGNIT aims to: 1) support an innovative new serverless paradigm for edge application management and enhanced digital sovereignty for users and developers; 2) enable on-demand deployment of large-scale, highly distributed and self-adaptive serverless environments using existing cloud resources; 3) optimize data placement according to changes in energy efficiency heuristics and application demands and behavior; 4) enable secure and trusted execution of serverless runtimes. We identify and discuss seven research challenges related to the integration of serverless technologies with multi-provider Edge infrastructures and present our vision for how these challenges can be solved. We introduce a high-level view of our reference architecture for serverless cloud-edge continuum systems, and detail four motivating real-world use cases that will be used for validation, drawing from domains within Smart Cities, Agriculture and Environment, Energy, and Cybersecurity.
With the inception of 6LoWPAN, it is possible to connect wireless sensor networks (WSN) and smart objects with the Internet using the IPv6 protocol, hence forming the IPv6-based Internet of Things (6IoT). Since the links in the 6IoT are lossy, UDP rather than TCP is mostly used for the communication between things. For the same reason, CoAP, a connection-less variant of HTTP, is being standardized as the web protocol for the 6IoT. Due to the sensitivity of the potential applications and presence of humans in the loop, End-to-End (E2E) security between constrained devices and hosts on Internet is one of the main requirements in the 6IoT. Secure CoAP (CoAPs) is used to provide end-to-end security in the CoAP-based 6IoT. Smartphones with sensing capabilities, direct human interaction, Internet connectivity, and relatively powerful processing and storage capabilities, are going to be an integral part of the 6IoT. In this paper we design, implement, and evaluate CoAPs for Android powered smartphones. We call our CoAPs INDIGO. To the best of our knowledge this is the first work that provides CoAPs support in smartphones. We implement and evaluate all cryptographic cipher suites proposed in the CoAP protocol, including the certificate-based authentication using Ecliptic Curve Cryptography (ECC). We also present novel application scenarios that are enabled by INDIGO on smartphones.
Wireless Sensor and Actuator Networks (WSNs) are distributed sensor and actuator networks that monitor and control real-world phenomena, enabling the integration of the physical with the virtual world. They are used in domains like building automation, control systems, remote healthcare, etc., which are all highly process-driven. Today, tools and insights of Business Process Modeling (BPM) are not used to model WSN logic, as BPM focuses mostly on the coordination of people and IT systems and neglects the integration of embedded IT. WSN development still requires significant special-purpose, low-level, and manual coding of process logic. By exploiting similarities between WSN applications and business processes, this work aims to create a holistic system enabling the modeling and execution of executable processes that integrate, coordinate, and control WSNs. Concretely, we present a WSN-specific extension for Business Process Modeling Notation (BPMN) and a compiler that transforms the extended BPMN models into WSN-specific code to distribute process execution over both a WSN and a standard business process engine. The developed tool-chain allows modeling of an independent control loop for the WSN.
Car manufacturers are noticing and encouraging a trend away from individual mobility, where a vehicle is owned and driven by one or only a few other persons, and towards shared-mobility concepts. That means that many different people use and have access to the same vehicle. An attacker disguised as a regular short-time user can use the additional attack vectors (s)he gets by having physical access to tamper the vehicle's software. The software takes a continuously more crucial role in cars for autonomous driving, and manipulations can have catastrophic consequences for the persons on board. Currently, there is no mechanism available to the vehicle owner to detect such manipulations in the vehicle done by the attacker (short-time user). In this work, a novel vehicle attestation scheme called Vehicular Soft Integrity Preservation Scheme (VeSIPreS) is proposed to detect tampering in the software stack of a vehicle and guarantee the upcoming driver that the previous user has not changed the software of the vehicle. The solution consists of a software module in the vehicle and a mobile-based user application for the vehicle owner to monitor the vehicle's soft integrity. Inside the vehicle, the software module is implemented in the central gateway, which acts as the primary security component. VeSIPreS uses Trusted Platform Module (TPM) in the central gateway, which anchors trust in our proposed solution. This paper also provides a proof-of-concept implementation with a TPM, demonstrating its application and deployment feasibility and presentig a security analysis to show the security of VeSIPreS.
Humans are social animals. They realize their dreams and ambitions in relation to other humans, within a context of symphonic diversity. They feel good or bad, mainly in relation to other people. And by doing it, we – the sensual humans – feel with the others, we feel for the others. I believe that the role of Design is to offer tools and ways to learn to deal in a creative, constructive way, how to be together, how to weave these relationships among people, how to play with and transform towards the resistance and the ambiguity that the environment and the people within this environment create, in their constant dynamism. By leveraging on people’s skills, and therefore on their sensitivity, and their urge of transformation, Design can concur in creating and consolidating practices, able to elicit ethical behaviors, building new foundations of social health. Design can elicit health by Making: less noise and more deeds, for a new craftsmanship.
Through the course Dense Spaces 2012—i.e. designing small, intelligent spaces such as elevators—carried out together with a group of architecture students at Umeå School of Architecture, Umeå University, Sweden, we report on, exemplify, and discuss how architectural theories, skills, and attitudes can come to complement and provide new food for thought for other design fields, including interaction design. We present the course, discuss some resulting spaces, and reflect on feedback from the participants. Then, we discuss some outcomes of the course that have broader implications. Unlike a more traditional technology-centered perspective, an architectural approach seems more prone to focus on designing what we term dynamic absence, i.e. design also concerned with what is not there. In a similar vein, an architectural approach also seems to address complexity by not fragmenting design challenges into smaller problems. The more holistic architectural attitude provides the opportunity to treat technology as a design material, along with the other architectural design materials the design situation offers, including structures, light, space, and absence. In this way, the architectural approach seems to shift the attention away from the design of representations and metaphors to instead focus on designing meaningful engagements in these spaces.
Potentialities of skills in design are intriguing. Skills open up new perceptions of the world, transform human understanding and engagement with the world itself. Our explorations suggest that leveraging existing designer's skills and training for new skills might remarkably contribute in designing for richness of meaning. We developed several skills-based techniques and validated them through a number of workshops. These techniques encourage participants to make before thinking, to reflect on the outcomes of making, and proceed by iterations of reflection-on-action. Also developed are techniques to increase the frequency of such iterations to minimize loss of meaning by abstraction, and techniques to foster depth of reflection. We organised these techniques into a framework, Designing in Skills (DiS). DiS nurtures personal engagement of designers, compelling a sense of responsibility; it supports designers toward what we call the "first-person perspective", enabling application of individual sensitivities. This paper presents firstly the motivation of our work and the surrounding theory. Subsequently, it introduces the framework and its development, using design cases that have led to its consolidation. It illustrates how DiS prepares for design practice and reflects on the theme of experiential richness.
This paper reflects on the possibilities of embodiment and skilful coping to connect people and to catalyse a constructive (design) “conversation” among people with different backgrounds, during transformative collaboration. We do this by illustrating the process and results of a two- weeks design class with Master students at the Department of Industrial Design at the Eindhoven University of Technology. The resulting Engagement Catalyser is a creative tool to engaging people in a (design) discussion more concrete and effective than a discussion or brainstorm session held around a table. The six developed Engagement Catalysers have been used and evaluated in two workshops, in which participants from very different cultural and professional background have used them as a means to engage quickly and ignite the design process. The results show that the Catalysers stimulate engagement, help people to get familiar and connected in a short period of time, and seem to inspire and boost the design process.
This video shows the ongoing design research project MoCap Tango. The project highlights the subtle qualities embedded in the physical dialogue between two tango dancers from a design perspective. Using custom-made wearables fitted with passive markers, in an optical Motion Capture System, the movements of two world-class tango dancers are captured. This data is used to experiment with real-time visualisations and 3D printed materialisations of the movements. The video presents the current state of the project, showing public performances in which the system was used as well as current work to use the data to create animations and 3D printed sculptures. Interviews with part of the design team highlight motivations for the project and discuss its relevance for embodied interaction design.
The +++ Wearable Player is a result of the application of the Rights through Making approach in designing wearables. This approach aims at designing systems, whose use empowers people towards the materialization of values (e.g. human rights). The +++ Wearable Player system elaborates on the previous project Sound Experience, and introduces the concept of viral music exchange as a motivating factor in the context of social health. This paper describes the morphological genesis, the functional aspects and how they have been implemented in a fully working experienceable prototype. The design process and its outcomes are illustrated, in the framework of the “changing behaviour” design trend.
As HCI becomes ever-increasingly more transdisciplinary it encounters increasingly complex problems practical, methodological, and pedagogical in natures. This paper is an introductory exploration of the influence HCI education has in bridging academia and industry as students become practitioners. We examined how design pedagogy materializes and takes shape in both work and student process/attitudes as they become professionals, suggesting there is an area of importance to the community that is overlooked. Education shapes designers, designers shape the world, which prompts the need for a dialogue on how education pedagogy shapes practitioners that embody methods, values, skills, goals, and practices. As practitioners embody their knowledge into designs there arises a discussion that ought to be had.
Distributed Hash Tables (DHTs) provide scalable mechanisms for implementing resource discovery services in structured Peer-to-Peer (P2P) networks. However, DHT-based lookups do not support some types of queries which are fundamental in several classes of applications. A way to support arbitrary queries in structured P2P networks is implementing unstructured search techniques on top of DHT-based overlays. This approach has been exploited in the design of DQDHT, a P2P search algorithm that combines the dynamic querying (DQ) technique used in unstructured networks with an algorithm for efficient broadcast over a DHT. Similarly to DQ, DQ-DHT dynamically adapts the search extent on the basis of the desired number of results and the popularity of the resource to be found. Differently from DQ, DQ-DHT exploits the structural constraints of the DHT to avoid message duplications. The original DQ-DHT algorithm has been implemented using Chord as basic overlay. In this paper we focus on extending DQ-DHT to work in k-ary DHT-based overlays. In a k-ary DHT, broadcast takes only O(logkN) hops using O(logkN) pointers per node. We exploit this “k-ary principle” in DQ-DHT to improve the search time with respect to the original Chord-based implementation. This paper describes the implementation of DQ-DHT over a k-ary DHT and analyzes its performance in terms of search time and generated number of messages in different configuration scenarios.
Software used in sensor networks to perform tasks such as sensing, network routing, and operating system services can be subject to changes or replacements during the long lifetimes of many sensor networks. The task of reprogramming the network nodes consumes a significant amount of energy and increases the network congestion because the software is sent over radio in an epidemic manner to all the nodes. In this thesis, I show that the use of data compression of dynamically linkable modules makes the reprogramming more energy-efficient and reduces the time needed for the software to propagate. The cost for using data compression is that the decompression algorithm must execute for a relatively long time, and that the software takes five to ten kilobytes to store on the sensor nodes.
Persistent storage offers multiple advantages for sensor networks, yet the available storage systems have been unwieldy because of their complexity and device-specific designs. We present the Coffee file system for flash-based sensor devices. Coffee provides a programming interface for building efficient and portable storage abstractions. Unlike previous flash file systems, Coffee uses a small and constant RAM footprint per file, making it scale elegantly with workloads consisting of large files or many files. In addition, the performance overhead of Coffee is low: the throughput is at least 92\% of the achievable direct flash driver throughput. We show that network layer components such as routing tables and packet queues can be implemented on top of Coffee, leading to increased performance and reduced memory requirements for routing and transport protocols.
The E-Care@Home Project aims at providing a comprehensive IoT-based healthcare system, including state-of-the-art communication protocols and high-level analysis of data from various types of sensors. With this poster, we present its novel technical infrastructure, consisting of low-power IPv6 networking, sensors for health monitoring, and resource-efficient software, that is used to gather data from elderly patients and their surrounding environment.
Low-power networked devices, such as sensors and actuators, are becoming a vital part of our everyday infrastructure. Being networked, the continued development of these systems needs involvement of the networking community. We present a framework for simulation, experimentation, and evaluation of routing mechanisms for low-power IPv6 networking. The framework provides a detailed simulation environment for low-power routing mechanisms and allows the system to be directly uploaded to a physical testbed for experimental measurements.
We present Velox, a virtual machine architecture that provides a safe execution environment for applications in resource-constrained IoT devices. Our goal with this architecture is to support developers in writing and deploying safe IoT applications, in a manner similar to smartphones with application stores. To this end, we provide a resource and security policy framework that enables fine-grained control of the execution environment of IoT applications. This framework allows device owners to configure, e.g., the amount of bandwidth, energy, and memory that each IoT application can use. Velox's features also include support for high-level programming languages, a compact bytecode format, and preemptive multi-threading.
In the context of IoT devices, there are typically severe energy, memory, and processing constraints that make the design and implementation of a virtual machine with such features challenging. We elaborate on how Velox is implemented in a resource-efficient manner, and describe our port of Velox to the Contiki OS. Our experimental evaluation shows that we can control the resource usage of applications with a low overhead. We further show that, for typical I/O-driven IoT applications, the CPU and energy overhead of executing Velox bytecode is as low as 1–5% compared to corresponding applications compiled to machine code. Lastly, we demonstrate how application policies can be used to mitigate the possibility of exploiting vulnerable applications.
Studies using Nomura et al.’s “Negative Attitude toward Robots Scale” (NARS) [1] as an attitudinal measure have featured robots that were perceived to be autonomous, independent agents. State of the art telepresence robots require an explicit human-in-the-loop to drive the robot around. In this paper, we investigate if NARS can be used with telepresence robots. To this end, we conducted three studies in which people watched videos of telepresence robots (n=70), operated telepresence robots (n=38), and interacted with telepresence robots (n=12). Overall, the results from our three studies indicated that NARS may be applied to telepresence robots, and culture, gender, and prior robot experience can be influential factors on the NARS score.
A learning system for fault finding has been constructed. This system contains many different types of knowledge, three ways to find faults and four ways to learn fault finding. The constructed learning system works for a class of fault finding problems. This class has been described in the paper. The developed system could be viewed as an architecture of a general learning system for fault finding. The system could also be used as a testbench of learning mechanisms. The experiences from this project indicates that it is possible to build a learning system when the structure of the knowledge is known. In this paper the following ideas will be discussed: How can the fault finding and learning techniques be integrated? How can the knowledge structure, fault finding mechanisms and learning mechanisms emerge with the help of a simulator and general mechanisms?
Multimodal conversational spoken dialogues using physical and virtual agents provide a potential interface to motivate and support users in the domain of health and fitness. In this paper we present how such multimodal conversational Companions can be implemented to support their owners in various pervasive and mobile settings. In particular, we focus on different forms of multimodality and system architectures for such interfaces.
As many countries are about to make changes in the primary school curriculum by introducing computational thinking, new methods and support for teachers is needed in order help them develop and adapt teaching materials. In this paper, technical pitfalls and other considerations for designing teaching materials with the microcontroller BBC micro:bit are presented. The results are based on a series of 21 workshops in different parts of Sweden aiming to investigate what is important to consider when designing teaching materials with the BBC micro:bit for training Swedish primary schools students computational thinking skills. The contribution of the paper are a number of identified considerations that can be helpful for teachers when designing exercises and planning for teaching computational thinking with the BBC micro:bit.Considerations and Technical Pitfalls for Teaching Computational Thinking with BBC micro:bit | Request PDF. Available from: https://www.researchgate.net/publication/326026189_Considerations_and_Technical_Pitfalls_for_Teaching_Computational_Thinking_with_BBC_microbit [accessed Aug 21 2018].
Subjective language is used to express attitudes and opinions towards things, ideas and people. While content and topic centred natural language processing is now part of everyday life, analysis of subjective aspects of natural language have until recently been largely neglected by the research community. The explosive growth of personal blogs, consumer opinion sites and social network applications in the last years, have however created increased interest in subjective language analysis. This paper provides an overview of recent research conducted in the area.
I describe a fast multilingual parser for semantic dependencies. The parser is implemented as a pipeline of linear classifiers trained with support vector machines. I use only first order features, and no pair-wise feature combinations in order to reduce training and prediction times. Hyper-parameters are carefully tuned for each language and sub-problem. The system is evaluated on seven different languages: Catalan, Chinese, Czech, English, German, Japanese and Spanish. An analysis of learning rates and of the reliance on syntactic parsing quality shows that only modest improvements could be expected for most languages given more training data; Better syntactic parsing quality, on the other hand, could greatly improve the results. Individual tuning of hyper-parameters is crucial for obtaining good semantic parsing quality.
In this paper, we study direct transfer methods for multilingual named entity recognition. Specifically, we extend the method recently proposed by Täckström et al. (2012), which is based on cross-lingual word cluster features. First, we show that by using multiple source languages, combined with self-training for target language adaptation, we can achieve significant improvements compared to using only single source direct transfer. Second, we investigate how the direct transfer system fares against a supervised target language system and conclude that between 8,000 and 16,000 word tokens need to be annotated in each target language to match the best direct transfer system. Finally, we show that we can significantly improve target language performance, even after annotating up to 64,000 tokens in the target language, by simply concatenating source and target language annotations.
Contemporary approaches to natural language processing are predominantly based on statistical machine learning from large amounts of text, which has been manually annotated with the linguistic structure of interest. However, such complete supervision is currently only available for the world's major languages, in a limited number of domains and for a limited range of tasks. As an alternative, this dissertation considers methods for linguistic structure prediction that can make use of incomplete and cross-lingual supervision, with the prospect of making linguistic processing tools more widely available at a lower cost. An overarching theme of this work is the use of structured discriminative latent variable models for learning with indirect and ambiguous supervision; as instantiated, these models admit rich model features while retaining efficient learning and inference properties. The first contribution to this end is a latent-variable model for fine-grained sentiment analysis with coarse-grained indirect supervision. The second is a model for cross-lingual word-cluster induction and the application thereof to cross-lingual model transfer. The third is a method for adapting multi-source discriminative cross-lingual transfer models to target languages, by means of typologically informed selective parameter sharing. The fourth is an ambiguity-aware self- and ensemble-training algorithm, which is applied to target language adaptation and relexicalization of delexicalized cross-lingual transfer parsers. The fifth is a set of sequence-labeling models that combine constraints at the level of tokens and types, and an instantiation of these models for part-of-speech tagging with incomplete cross-lingual and crowdsourced supervision. In addition to these contributions, comprehensive overviews are provided of structured prediction with no or incomplete supervision, as well as of learning in the multilingual and cross-lingual settings. Through careful empirical evaluation, it is established that the proposed methods can be used to create substantially more accurate tools for linguistic processing, compared to both unsupervised methods and to recently proposed cross-lingual methods. The empirical support for this claim is particularly strong in the latter case; our models for syntactic dependency parsing and part-of-speech tagging achieve the hitherto best published results for a wide number of target languages, in the setting where no annotated training data is available in the target language.
We present an application of transductive semi-supervised learning to the problem of speaker identification. Formulating this problem as one of transduction is the most natural choice in some scenarios, such as when annotating archived speech data. Experiments with the CHAINS corpus show that, using the basic MFCC-encoding of recorded utterances, a well known simple semi-supervised algorithm, label spread, can solve this problem well. With only a small number of labelled utterances, the semi-supervised algorithm drastically outperforms a state of the art supervised support vector machine algorithm. Although we restrict ourselves to the transductive setting in this paper, the results encourage future work on semi-supervised learning for inductive speaker identification.
We have investigated the potential for improvement in target language morphology when translating into Swedish from English and German, by measuring the errors made by a state of the art phrase-based statistical machine translation system. Our results show that there is indeed a performance gap to be filled by better modelling of inflectional morphology and compounding; and that the gap is not filled by simply feeding the translation system with more training data.
This paper presents work in progress on implementing an embodied question answering system, Dr. Cecilia, in the form of a virtual caregiver, for use in the treatment of eating disorders. The rationale for the system is grounded in one of the few effective treatments for anorexia and bulimia nervosa. The questions and answers database is encoded using natural language, and is easily updatable by human caregivers without any technical expertise. Matching of users' questions with database entries is performed using a weighted and normalized n-gram similarity function. In this paper we give a comprehensive background to and an overview of the system, with a focus on aspects pertaining to natural language processing and user interaction. The system is currently only implemented for Swedish.