Selective foregrounding of contextual elements as a form of message enrichment in advertising

This paper approaches the use of both context and semantic information in the information retrieval process with the goal of developing context-based semantically enhanced information retrieval systems. To achieve our objective we have identified, defined and formalized three distinct types of context information relevant for an information retrieval system: knowledge context information, user context information and constraint context information. The context information is represented in an information system interpretable way by mapping it onto our RAP context model elements. The proposed information retrieval model is tested using the arhiNet system, our integrated information retrieval system for archive content, based on semantic enhancements.

A probabilistic ontology-based platform for self-learning context-aware healthcare applications

The emergence of new types of mobile and embedded computing devices and developments in wireless networking are broadening the domain of computing from the workplace and home office to other facets of everyday life. This trend is expected to lead to a proliferation of pervasive computing environments, in which inexpensive, interconnected computing devices are ubiquitous and capable of supporting users in a range of tasks. It is widely accepted that the success of pervasive computing technologies will require a radical design shift, and that it is not sufficient to simply extrapolate from existing desktop computing technologies. In particular, pervasive computing demands applications that are capable of operating in highly dynamic environments and of placing fewer demands on user attention. In order to meet these requirements, pervasive computing applications need to be sensitive to the context of use, including the location, time and activities of the user. Currently, the programming of context-aware applications represents a complex and error-prone task, while modification to support changing user requirements or a changing set of context information is usually prohibitively difficult. Consequently, context-aware applications are explored largely in laboratory settings, and remain some distance from widespread ac ceptance and use. In order to remedy this situation, there is a need for better understanding of the design process associated with context-aware applications, improved programming models that lead to highly flexible and customisable applications, and infrastructural support for tasks such as gathering and management of context information. This thesis presents a framework that addresses these issues. The framework integrates a set of original conceptual foundations, including context and preference modelling techniques, with a software architecture that implements context and preference management functions and provides programming support in the form of a toolkit. The thesis makes several important research contributions. First, it presents a novel characterisation of context information in pervasive computing systems, covering (among other features) temporal aspects and various types and sources of uncertainty. Second, it proposes two complementary approaches to context modelling. The first modelling approach, CML, provides a graphical notation that supports the exploration and specification of an application's context requirements by the designer. CML represents context information in terms of facts, and has a strong formal basis that enables a straightforward mapping to a context management system built around a relational database. The second approach, termed the situation abstraction, allows contexts to be described in selective, high-level terms as constraints upon a fact-based CML model. Situations are well suited for use in context querying and as programming abstractions. Third, the thesis presents a pair of programming models that can be used in conjunction with the situation abstraction. The first model, which enables the triggering of actions in response to context changes, has been widely used previously in the development of adaptive and context-aware software, but is reformulated here to accommodate uncertain context information. The second model, which supports choice amongst alternative actions based on the context and preferences of the user (termed branching), is unique to this thesis, and is developed in conjunction with a novel preference modelling approach that allows users to easily express and combine context-dependent requirements. Fourth, the thesis proposes a software architecture for context-aware systems, which combines toolkit support for the two programming models with software components that perform gathering and processing of context information from a variety of sources, and management of both context and preference information. Finally, the thesis presents a case study that evaluates a partial implementation of the architecture and its underlying conceptual foundations. This involves the development of a context-aware communication platform that supports choice of communication channels for interactions between users based on the contexts and preferences of the participants. The case study validates the architecture, the context and preference modelling approaches and the branching model, and illustrates the process and issues involved in the design of context-aware software.

Rapid advances in computing technology have made it possible to create a pervasive computing environment where computing resources and facilities are available wherever a user may be. This pervasive computing environment goes beyond the concept of city-wide, or nation-wide, wireless network coverage and aims to make interaction between the user and the computing facilities as easy and natural as possible. This requires intelligence on the part of the underlying networks and software supporting the pervasive computing environment, and also awareness of the environment and the context in which the user is operating. Intelligence and awareness on the part of the pervasive computing environment are provided by analysing what is referred to as “context information”. This context information can be obtained from hardware or software sensors in the environment, be based on profiled information or derived from context information already known to the pervasive computing environment. The potential to misuse the pervasive computing infrastructure to violate the privacy of users is significant. These violations can range from trading context information about users without consent to abusing a user’s location information to stalk them. To make the link between users and context information gathered by sensors we developed a concept of contextdependent, multi-party context information ownership. In our approach, ownership of context information entitles the owner(s) to determine how, to whom and under what conditions context information is disclosed to third-party entities. This concept of ownership was then developed into modelling techniques and a representation compatible with a wide range of Object Oriented context modelling approaches. Context information privacy for the users of the pervasive computing environment was then addressed by building on our concept of context-dependent, multi-party context information ownership. Owners of context information, identified using our ownership modelling approach, were able to specify their disclosure preferences for context information modelled as belonging to them. These preferences were captured using a context-dependent privacy preference language. As part of the contribution to the field of context information privacy a context information obfuscation mechanism was also developed. Control of this obfuscation mechanism was integrated with the preference language permitting users to specify not only when, how, why and for what purpose their context information can be disclosed by the pervasive computing environment, but also at what level of detail disclosure may take place. If necessary, the obfuscation mechanism operates on requested context information to reduce its detail level to meet these disclosure requirements. After developing a context information privacy mechanism, we then focused on enhancing the operation of security mechanisms within pervasive computing environments by using context information. From a security perspective, the pervasive computing environment represents a significant challenge as (1) there is no single administrative entity to establish and enforce security policy, (2) nodes are highly mobile and may need to interact with other nodes with which they have no previous relationship and (3) connectivity may be intermittent meaning that trusted infrastructure may not always be available. These features of pervasive computing environments mean that existing approaches to security that rely on the existence of trusted infrastructure, need well-defined network boundaries and require pre-existing relationships between the security service are unable to function. As the security field is very broad, this thesis focused on one aspect: authentication. To facilitate authentication in pervasive computing environments, we developed the Distributed Certification and Authentication Service (DCAS). DCAS acts as a distributed Certification Authority comprised of a group of trustworthy nodes that collaborate to provide the services of a Certification Authority and can be set up on-the-fly to service clusters of isolated nodes. The trusted nodes within DCAS not only issue public key certificates, but also perform authentication of applicant nodes. Applicant nodes are able to authenticate using “traditional authentication factors” (such as public key certificates), as well as context information sensed from the environment, or a combination of both. To demonstrate the viability of DCAS, its core functionality was implemented and deployed on a small IEEE 802.11b network. In order to aid in establishing a DCAS within a network, a probability-based model of DCAS behaviour under certain conditions was created. This model was verified using simulation.

Context is critical for reducing the uncertainty in object detection. However, context modeling is challenging because there are often many different types of contextual information coexisting with different degrees of relevance to the detection of target object(s) in different images. It is therefore crucial to devise a context model to automatically quantify and select the most effective contextual information for assisting in detecting the target object. Nevertheless, the diversity of contextual information means that learning a robust context model requires a larger training set than learning the target object appearance model, which may not be available in practice. In this work, a novel context modeling framework is proposed without the need for any prior scene segmentation or context annotation. We formulate a polar geometric context descriptor for representing multiple types of contextual information. In order to quantify context, we propose a new maximum margin context (MMC) model to evaluate and measure the usefulness of contextual information directly and explicitly through a discriminant context inference method. Furthermore, to address the problem of context learning with limited data, we exploit the idea of transfer learning based on the observation that although two categories of objects can have very different visual appearance, there can be similarity in their context and/or the way contextual information helps to distinguish target objects from nontarget objects. To that end, two novel context transfer learning models are proposed which utilize training samples from source object classes to improve the learning of the context model for a target object class based on a joint maximum margin learning framework. Experiments are carried out on PASCAL VOC2005 and VOC2007 data sets, a luggage detection data set extracted from the i-LIDS data set, and a vehicle detection data set extracted from outdoor surveillance footage. Our results validate the effectiveness of the proposed models for quantifying and transferring contextual information, and demonstrate that they outperform related alternative context models.

In smart living spaces, a reliable context-aware application must adapt to the variable environment and should function to complete the user's requirements. Therefore, if a context-aware application is to achieve the user's requirements smoothly, it must understand the environment status according to trustworthy context information. However, the context information influenced by the environment status may be imperfect. If a context-aware application responds to the user's requirements using the incorrect context information, this will lead to undesirable result. To resolve these problems, this paper proposes a Context Model (CM) to evaluate the reliability of context information and predicted the context information that context-aware application needs. When CM approves of the context information, it considers both the key features of context information, and also the different requirements of the context-aware application. Finally, an experiment is proposed to verify the reliability of CM. The experiment verifies that the reliability of context information recommended by CM can represent the real status. In addition, the experiment verifies whether CM can provide appropriate context information according to the particular requirements of a context-aware application.

The notion of using context information for solving high-level vision and medical image segmentation problems has been increasingly realized in the field. However, how to learn an effective and efficient context model, together with an image appearance model, remains mostly unknown. The current literature using Markov Random Fields (MRFs) and Conditional Random Fields (CRFs) often involves specific algorithm design in which the modeling and computing stages are studied in isolation. In this paper, we propose a learning algorithm, auto-context. Given a set of training images and their corresponding label maps, we first learn a classifier on local image patches. The discriminative probability (or classification confidence) maps created by the learned classifier are then used as context information, in addition to the original image patches, to train a new classifier. The algorithm then iterates until convergence. Auto-context integrates low-level and context information by fusing a large number of low-level appearance features with context and implicit shape information. The resulting discriminative algorithm is general and easy to implement. Under nearly the same parameter settings in training, we apply the algorithm to three challenging vision applications: foreground/background segregation, human body configuration estimation, and scene region labeling. Moreover, context also plays a very important role in medical/brain images where the anatomical structures are mostly constrained to relatively fixed positions. With only some slight changes resulting from using 3D instead of 2D features, the auto-context algorithm applied to brain MRI image segmentation is shown to outperform state-of-the-art algorithms specifically designed for this domain. Furthermore, the scope of the proposed algorithm goes beyond image analysis and it has the potential to be used for a wide variety of problems for structured prediction problems.

We consider the problem of predicting instantiated binary relations in a multi-relational setting and exploit both intrarelational correlations and contextual information. For the modular combination we discuss simple heuristics, additive models and an approach that can be motivated from a hierarchical Bayesian perspective. In the concrete examples we consider models that exploit contextual information both from the database and from contextual unstructured information, e.g., information extracted from textual documents describing the involved entities. By using low-rank approximations in the context models, the models perform latent semantic analyses and can generalize across specific terms, i.e., the model might use similar latent representations for semantically related terms. All the approaches we are considering have unique solutions. They can exploit sparse matrix algebra and are thus highly scalable and can easily be generalized to new entities. We evaluate the effectiveness of nonlinear interaction terms and reduce the number of terms by applying feature selection. For the optimization of the context model we use an alternating least squares approach. We experimentally analyze scalability. We validate our approach using two synthetic data sets and using two data sets derived from the Linked Open Data (LOD) cloud.

In the present age, context-awareness is an important aspect of the dynamic environments and the different types of dynamic context information bring new challenges to access control systems. Therefore, the need for the new access control frameworks to link their decision making abilities with the context-awareness capabilities have become increasingly significant. The main goal of this research is to develop a new access control framework that is capable of providing secure access to information resources or software services in a context-aware manner. Towards this goal, we propose a new semantic policy framework that extends the basic role-based access control (RBAC) approach with both dynamic associations of user-role and role-service capabilities. We also introduce a context model in modelling the basic and high-level context information relevant to access control. In addition, a situation can be determined on the fly so as to combine the relevant states of the entities and the purpose or user’s intention in accessing the services. For this purpose, we can propose a situation model in modelling the purpose-oriented situations. Finally we need a policy model that will let the users to access resources or services when certain dynamically changing conditions (using context and situation information) are satisfied.

With the development of electronic and networking technologies, ubiquitous computing is now an exploring focused area. Due to characteristics of ubiquitous computing, privacy and security issues are always quite challenging for application development. This paper presents a novel paradigm to enhance the privacy and security in ubiquitous environments. Unlike previous context model, the proposed composite sensor-based context model (ComSensor) captures both complicated context entities and security requirements. Security agent governs security and privacy by privacy rules defined in ComSensor. One security agent is responsible for privacy management for one context objects. Sensitive context raw data and information are kept only in distributed security agents, from which context queries are processed and context information is disclosed in different accurate levels according to privacy rules. Based on ComSensor and security agent, decentralized trust management architecture is proposed, which is appropriate for different kinds of ubiquitous applications. Examining results on a prototype show that this paradigm is scalable and runs in real-time mode.

It has been shown in the literature that automatic speech recognition systems can greatly benefit from contextual information [1, 2, 3, 4, 5]. Contextual information can be used to simplify the beam search and improve recognition accuracy. Types of useful contextual information can include the name of the application the user is in, the contents of the user's phone screen, the user's location, a certain dialog state, etc. Building a separate language model for each of these types of context is not feasible due to limited resources or limited amounts of training data. In this paper we describe an approach for unsupervised learning of contextual information and automatic building of contextual biasing models. Our approach can be used to build a large number of small contextual models from a limited amount of available unsupervised training data. We describe how n-grams relevant for a particular context are automatically selected as well as how an optimal size of a final contextual model is chosen. Our experimental results show great accuracy improvements for several types of context.

A survey of context modelling and reasoning techniques

DGC: Dynamic group behavior modeling that utilizes context information for group recommendation

Defining a Context Model for Smart Manufacturing

Context-aware functionalities are functionalities that consider the context to produce a certain system behavior, typically an adaptation or recommendation. As contextual elements such as time, location, weather, user activity, device characteristics, network status, and countless others are becoming increasingly more accessible, the potential for adding context awareness to applications is enormous. Identifying novel, unexpected, and even delightful context-aware functionalities in practice can be challenging, though: What context information is relevant for a given user task? How can contextual elements be combined? What if there is a large number of contextual elements? Context modeling has been described in the literature as an essential aspect in the elicitation of context-aware functionalities; however, reports on the state of the practice are rare. In this study, we conducted a survey with industrial practitioners, mostly experienced professionals from large enterprises, to investigate how context models and context-modeling activities have been used to support the elicitation of context-aware functionalities. The results indicate a gap between research and industry: Context models are rarely used in practice, and context-modeling activities such as analysis of relevance and especially analysis of combinations of contextual elements have been overlooked due to their high complexity, despite practitioners recognizing their importance.