Modeling and Using Context

Despite recognition that context dependence is endemic to natural language, arguably one of its core properties, standard grammar formalisms remain poor vehicles for expressing this. Grammars are standardly defined in terms of principles underpinning structures inhabited by sentence strings to yield a concept of sentence wellformedness, these being defined independently of any attribute of performance; and all data not characterisable in these terms are seen as in some sense peripheral, requiring grammar-external explication. Phenomena displaying context dependence all pose problems for this methodology, context dependence being no respecter of sentence boundaries. Ellipsis poses this challenge in a particularly acute way. A reasonable goal for accounts of ellipsis might be seen as one providing formal explication of the informal insight that ellipsis constitutes the license for expressions to be omitted if recoverable from context, with ellipsis data taken as a window on the requisite concept of context for language construal. But such a goal isn’t even addressed as a desideratum in current accounts. Ellipsis is seen as falling into two major types: sentence-internal and discourse ellipsis; and even within the former type, there is further bifurcation into those types of ellipsis characterisable as semantically grounded (Dalrymple et al 1991) and those requiring syntactic characterisation with a number of discrete structural types (Merchant 2009). And, with a pragmatic remainder falling outside grammar-based characterisation (Stainton 2006), the phenomenon is seen as irreducibly “fractally heterogeneous” (Ginzburg and Cooper 2004).

Mobile phone games are played in context. Although such information has been used in several prototypes, very few context-aware games have made it beyond the research lab. In our research, we investigate how the development of context-aware games needs to be changed such that their commercialization is more feasible and they can be deployed more easily. Based on the findings of the creation and evaluation of a context-based game called

ContextSnake

, we developed a platform named

Gatherer

which frees the developer from the burden of collecting, preprocessing, storing, and interpreting raw sensor data. We introduce the novel concept of generic context which enables the use of context in mobile applications without having detailed information about the actual environment in which the system will be deployed. In order to preliminarily validate the platform, a second game called

ContextInvaders

developed on top of this platform is described.

Providing adequate support for context acquisition and management is a non-trivial task that complicates application development. To mitigate this, existing middleware systems rely on a centralized context service that manages the context of all devices located within a certain area. However, in many cases, centralized context management impacts privacy as it requires users to share their context with a possibly untrusted device. In this paper, we show how this problem can be avoided by a peer-based approach to context management. To validate the approach, we have implemented it on top of the BASE middleware and used it extensively in the PECES European research project. The evaluation shows that given a customizable implementation, the approach provides high flexibility while being suitable for a broad spectrum of devices.

In this paper we motivate and describe a dialogue manager (called

Frolog

) which uses classical planning to infer causal implicatures. A causal implicature is a type of Gricean relation implicature, a highly context dependent form of inference. As we shall see, causal implicatures are important for understanding the structure of task-oriented dialogues. Such dialogues locate conversational acts in contexts containing both pending tasks and the acts which bring them about. The ability to infer causal implicatures lets us interleave decisions about “how to sequence actions” with decisions about “when to generate clarification requests”; as a result we can model task-oriented dialogue as an interactive process locally structured by negotiation of the underlying task. We give several examples of

Frolog

-human dialog, discuss the limitations imposed by the classical planning paradigm, and indicate the potential relevance of our work for other relation implicatures.

Scientific-Workflow (SWF) management is similar to practice management. However, SWFs are stored in a repository as independent items that are reused by other actors after a complex process of contextualization, decontextualization and recontextualization. Conversely, practices are contextualized procedures applicable in different contexts. SWFs are collected in repositories in a flat way independently of each other, while practices can be organized in a uniform representation of elements of knowledge, reasoning and contexts. This paper shows that SWF management may benefit of the contextual approach used for representing procedures and practices. We propose a three-step approach, namely a support to scientist for finding the right SWF in the repository, a generation of all the possible SWFs in a situation, and the interactive development of SWFs. Thus, a SWF system will be flexible enough to acquire new knowledge incrementally and learn new ways for SWF building from scientists. A challenge for the short term is the possibility to model cooperation by the coupling of actors’ task representation through a shared context.

As embedded sensing systems are central to developing pervasive, context-aware services, the applications running on these systems should be intelligible to system programmers and to users. Given that sensor systems are programmed in low-level languages, manually writing high-level explanations about their decision model requires knowledge about the system architecture, and is error-prone. We explore the possibility of extracting explanations which are small and expressive, but still preserve bit-level accuracy when needed. We contribute a tool which automatically and soundly generates compact, graphical explanations from sensor software implementation at compile-time. We base our algorithm on the techniques of (i) finite-state machine model extraction from software as used in model checking, and (ii) abstraction of program execution traces. We experiment with extracting explanations from heavyweight, low-level TinyOS applications for a mainstream sensor platform.

In this paper we present a

business

context

information

manager

based on a novel and generic interpretation of the context. This manager takes into account various contextual dimensions and acts as an intermediary between information system and contextual information. First, we discuss how the manager creates a snapshot of the context that is provided to the information system for different purposes. Then, we introduce a process (MES) that manages the various contextual dimensions using a rule set to create a unique situation at a moment

t

. These rules are part of the knowledge hold by the context manager. The situation inference is particularly related to the interactions between contextual dimensions. Thus the knowledge can evolve owing to an extracting and learning process which improves the context manager reliability.

The paper examines the determination of the explicit content of communication via inferential processes based on the speakers’ situational interests. The paper’s argument is based on the hypothesis that the meaning of natural language expressions depends on the situational extra-linguistic interests of speakers. It is maintained that interests affect the sentences of natural language when semantic conventions, semantic context, and pragmatic context are unable to determine a unique meaning for a sentence. In particular, the paper addresses the question of the determination of meaning via inference where the speaker’s interest is a premise for communicative inferences. The notion of interest is viewed as (preference for) a state of affairs which implies the possibility conditions of the agent’s goal. The context of interests is viewed as the set of states of affairs that make a sentence true and contemporaneously satisfy the situational interests of the speaker. The last part of the paper illustrates a real-life case of linguistic dispute.

In pervasive environments, presence-based application development via Presence Management Systems (PMSs) is a key factor to optimise the management of communication channels, driving productivity increase. Solutions for presence management should satisfy the interoperability requirements, in turn providing context-centric presence analysis and privacy management. In order to push PMSs towards flexible, open and context-aware presence management, we propose some adaptation of two extensions to standard XML-based XMPP for message exchange in online communication systems. The contribution allows for more complex specification and management of nested group and privacy lists, where semantic technologies are used to map all messages into RDF vocabularies and pave the way for a broader semantic integration of heterogeneous and distributed presence information sources in the standard PMSs framework.

To enable deep learning from a system is not a trivial matter. By including domain context in the construction of the explanations the user is able to reflect on the effects of the knowledge, and not solely on the knowledge in isolation. In addition, in our approach the contextualized knowledge is to be viewed in the context of a particular goal or objective and how any reasoning result is reached in relation to this objective. In this paper we outline context-based explanations and their composition for a sample explanation supporting teachers in developing their own didactical skills.

Pervasive computing envisions implicit interaction between people and their intelligent environments instead of individual devices, inevitably leading to groups of individuals interacting with the same intelligent environment. These environments must therefore be aware not only of user contexts and activities, but the contexts and activities of groups of users as well. This poster will demonstrate an experiment conducted towards understanding hierarchical multi-user group activity recognition using wearable sensors. The experiment will explore different data abstraction levels in terms of recognition rates, power consumption and wireless communication volumes for the devices involved.

This paper examines global context classification in peer-to-peer ad-hoc mobile wireless networks (P2P-MANETs). To begin, circumstances are presented in which such systems would be required to classify a global context. These circumstances are expounded upon by presenting concrete scenarios from which a set of requirements are derived. Using these requirements, related work is evaluated for applicability, indicating no adequate solutions. Algorithmic approaches are proposed, and analysis results in a benchmark as well as bounds for distribution of processing load, memory consumption and message passing in P2P-MANETs.

Describing context for using it within a research and development process is an ambitious task. Different people with different background knowledge, goals, and thus aspects on context are participating in such a process and require context information for different purposes. In this paper we address this issue by defining a structural description model where we regard factors relevant to the different participants and thus perspectives of such a process. The proposed model is based on context models existing in the literature and current ways of using context in the areas of software engineering and human computer interaction.

Affective science conducts interdisciplinary research into the emotions and other affective phenomena. Currently, such research is hampered by the lack of common definitions of terms used to describe, categorise and report both individual emotional experiences and the results of scientific investigations of such experiences. High quality ontologies provide formal definitions for types of entities in reality and for the relationships between such entities, definitions which can be used to disambiguate and unify data across different disciplines. Heretofore, there has been little effort directed towards such formal representation for affective phenomena, in part because of widespread debates within the affective science community on matters of definition and categorization. To address this requirement, we are developing an Emotion Ontology (EMO). The full ontology and generated OWLDoc documentation are available for download from https://emotion-ontology.googlecode.com/svn/trunk/ under the Creative Commons – Attribution license (CC BY 3.0).

Ramsey’s test for conditionals seems to be in conflict with the so-called Thomason conditionals. A Thomason conditional is a conditional in which either the antecedent or the consequent is a statement about the reasoning agent’s own beliefs. Several authors have pointed out that resolving the apparent conflict is to be sought by abandoning the belief revision interpretation of the Ramsey test in favor of a suppositional interpretation. We formalize an AGM-style notion of supposition, showing that it is identical to revision for agents who are not autodoxastic—agents who do not reason about their beliefs. We present particular realizations of supposition in terms of revision and identify the relations between the conditionals supposition and revision give rise to.

Spiking neural networks are promising candidates for representing aspects of cognitive context in human memory. We extended the liquid state machine model with time-delayed connections from liquid neurons to the readout unit to better capture context phenomena. We performed experiments in the area of spoken language recognition for studying two aspects of context dependency: influence of memory and temporal context. For the experiments, we derived a test data set from the well-known Brody-Hopfield test set to which we added varying degrees of Gaussian noise. We studied the influence of temporal context with a further specially designed test set. We found that the temporal context encoded in the pattern to be recognized was recognized better with our delayed synapses than without. Our experiments shed light on how context serves to integrate information and to increase robustness in human signal processing.

When designing and implementing real world ambient intelligent systems we are in need of applicable information systems engineering methods. The tools we find in the intelligent systems area focus on the knowledge engineering parts, whereas traditional software engineering techniques are usually not designed with the peculiarities of intelligent systems design in mind. This holds in particular for explanation-aware intelligent systems. This work looks at problem frames for explanations and investigates how problem frames can be used to elicit, analyse, and specify these specific requirements. The point of departure is an existing ambient intelligent information system for the hospital ward domain. The work presented here analyses how such a system can be redesigned with a focus on explanation-awareness.

Context aware applications are reactive, they adapt to an entity’s context when the context has changed. In order to become proactive and act before the context actually has changed future contexts have to be predicted. This will enable functionalities like preloading of content or detection of future conflicts. For example if an application can predict where a user is heading to it can also check for train delays on the user’s way. So far research concentrates on context prediction algorithms that only use a history of one context to predict the future context. In this paper we propose a novel multidimensional context prediction algorithm and we show that the use of multidimensional context histories increases the prediction accuracy. We compare two multidimensional prediction algorithms, one of which is a new approach; the other was not yet experimentally tested. In theory, simulation and a real world experiment we verify the feasibility of both algorithms and show that our new approach has at least equal or better reasoning accuracy.

This paper presents a novel linguistic information extraction approach exploiting analysts’ stock ratings for statistical decision making. Over a period of one year, we gathered German stock analyst reports in order to determine market trends. Our goal is to provide business statistics over time to illustrate market trends for a user-selected company. We therefore recognize named entities within the very short stock analyst reports such as organization names (e.g. BASF, BMW, Ericsson), analyst houses (e.g. Gartner, Citigroup, Goldman Sachs), ratings (e.g. buy, sell, hold, underperform, recommended list) and price estimations by using lexicalized finite-state graphs, so-called local grammars. Then, company names and their acronyms respectively have to be cross-checked against data the analysts provide. Finally, all extracted values are compared and presented into charts with different views depending on the evaluation criteria (e.g. by time line). Thanks to this approach it will be easier and even more comfortable in the future to pay attention to analysts’ buy/sell signals without reading all their reports.

This paper describes a toolkit for implementing lifelog applications that automatically generate web-based reports about daily lives from sensor data such as GPS data, acceleration data, and camera images that capture our daily activities. To design the toolkit, we collect example lifelog applications using a questionnaire survey where we ask participants what kinds of daily life data they want to record for later use and then clarify their characteristics. Based on the obtained information and our view, we realize an infrastructure that provides automatic report publishing and development support for generating reports from sensor data.

The typical approach taken by clinical studies examining the factors that affect mood is to use questionnaires in order to record the activities that impact the mood. However, recording activities in this manner suffers from a number of issues including floor effect and difficulty in recalling past activities. Our work instead has focused on using unobtrusive monitoring technology to study mood changes during office hours and two associated factors that influence these changes, namely social activity and non-sedentary patterns. The pilot study ran over the course of 7 days of measurements with the participation of 9 knowledge workers. The results have shown that mood changes are highly correlated with both social interactions and non-sedentary work style. This study is the first to investigate the correlation between mood changes and non-sedentary behavior patterns, opening up a research avenue to explore psychological effects of increasing prevalence of sedentary behavior.

Digital Preservation is an important challenge for the information society. Reliable information and communication technology is crucial for most companies and software failure, is a considerable risk. Use of technologies such as Software as a Service (SaaS) and Internet of Services (IoS) means that business processes are increasingly supported by distributed, service oriented systems. We propose a concept and methods for capturing of contextual information, event causality and timing for Digital Preservation of distributed business processes and services. The architecture is derived from an architecture for monitoring sensing systems. We add a reasoner that can check whether processes adhere to explicit contracts and detect behavior anomalies, and we sketch how an inductive learner can be used to detect anomalies not covered by these contracts.

This article presents i-collaboration 3.0, a framework that supports the promotion of personalized distributed collaboration between students that collaborate to learn by using their Web 2.0 tools of choice (MSN, Twitter, Facebook,...). The framework is an extension of i-collaboration, a conceptual model that used Virtual Learning Companions (VLC) and the MBTI personality test to personalize learners’ experience in VLE. I-Collaboration 3.0 has a conceptual model that includes the pedagogical objective of the framework, the VLC new responsibilities in the framework and the students’ contextual information modeling in order to ensure a more interactive experience for the Learner. Moreover, it counts with an infrastructure already implemented that is compliant with the model and that includes the integration of the framework with Twitter (implementation of VLC as a Twitter user, where students can make questions by @mentions), MSN (implementation of VLC as a MSN user, where students can add him as a MSN contact and make questions) and Websites (implementation of VLC as a chatterbot - can be integrated with any website or blog).

We propose an in-home living activity recognition method using only off-the-shelf sensors, namely, an accelerometer and a microphone, which are commonly applied in mobile phones. The proposed method firstly estimates a user’s movement condition roughly by acceleration sensing. Secondly, it classifies the working condition in detail by acoustic sensing when it estimates the condition to be working by acceleration sensing. We developed a prototype system to recognize the user’s living activity in real time and conducted two experiments to confirm the feasibility of the proposed method. As a result of the first experiment, three movement conditions; quiet, walking, and working, are classified with more than 95% accuracy by acceleration sensing. And it classified working into seven conditions with 85.9% accuracy by acoustic sensing. Moreover, the result of the second experiment shows that it is effective to adopt instance-based recognition according to the assumed application.

Web search behavior is being diversified. And the diversification of web search behavior makes query both sporadic and non-sequential. In this case, it is difficult to understand why and how users search any information. Context of web search makes it possible to classify sporadic and non-sequential queries according to each interest. However, only the user him/herself can identify the context of web search behavior exactly. Hence, it is necessary to conduct a client-side research in which users are deeply engaged. The purpose of this research is to develop and apply the methodology that systematically examines the web search behavior and its context. To achieve this, (1) client-side log data is collected, then participants (2) clustered the queries based on context and (3) filled in a questionnaire as to each clustered queries. Also, interviews were conducted in each person. The finding of this study is that the features of UCQS are different from previous studies. Furthermore, we identified the flow of users’ web search behavior.

Typical localization systems use various features of the signal to estimate the distance, including received signal strength indicator (RSSI), timing information or angle of arrival (AoA). However, there are a number of signal features of FM radio that may also be suitable for localization, namely stereo channel separation (SCS) and signal to noise ratio (SNR). This paper investigates the feasibility of indoor localization using fingerprinting of audio features of FM radio signals emitted by low-power FM transmitters using SNR and SCS values. The experimental results demonstrate the possibility of audio-based localization, when signal strength readings are not available.

Modern smartphones provide sensors that can be used to describe the current context of the device and its user. Contextual knowledge allows software systems to adapt to personal preferences of users and to make data processing context-aware. Different sensors or measurement approaches used for recognizing the values of particular context elements vary greatly in their energy consumption. This paper presents approaches for reducing the energy consumption of utilizing smartphone sensors. We discuss sensor substitution strategies as well as logical dependencies among sensor measurements. The paper describes the first milestone towards a generalization of such strategies. Furthermore, We show that energy awareness benefits from a more abstract view on context elements.

Location-aware systems are mobile or spatially distributed computing systems, such as smart phones or sensor nodes in wireless sensor networks, enabled to react flexibly to changing environments. Due to severe restrictions of computational power on these platforms and real-time demands, most current solutions do not support advanced spatial reasoning. Qualitative Spatial Reasoning (QSR) and granularity are two mechanisms that have been suggested in order to make reasoning about spatial environments tractable. We propose an approach for combining these two techniques, so as to obtain a light-weight QSR mechanism, called

partial order QSR

(for brevity: PQSR), that is fast enough to allow application on small, low-cost computing devices. The key idea of PQSR is to use a core fragment of typical QSR relations, which can be expressed with partial orders and their linearizations, and to additionally delimit reasoning about these relations with a size-based granularity mechanism.

This paper outlines our current research program in the fields of ambient intelligence and context-aware computing and the tools we are building to accomplish this research program. From a discussion of our conception of mental models in the domain of ambient context-aware computer systems we derive hypotheses which we intend to test empirically. A modular framework for implementing and assessing situation awareness in humans and computers is introduced. We describe the framework’s architecture and illustrate its suitability for its intended purpose. Finally, we present an outline of our next steps towards real world application systems for our research.

Knowledge intensive work is characterized by dealing with an increasing amount of resources, like documents or e-mails. To increase efficiency, users frequently reuse existing resources, e.g., create new documents by using existing ones as a template. This paper introduces

LiCoRMS

, a lightweight system for supporting the reuse of resources by capturing and managing relationships between resources using lifecycle

and

content information in resource-centered metadata as well as in an ontology.

Context-based authentication methods enable the unobtrusive establishment of authentication or even secure keys. While several context-based authentication methods have been proposed recently, often the entropy of the seed for the cryptographic keys is not exploited. We study the entropy of audio fingerprints which can be utilized to pair devices in close proximity. In this work, for 600 audio fingerprints from five distinct audio classes recorded at three different locations, we applied 7490 statistical tests from the dieHarder battery of statistical tests.

Most multiagent systems (MAS) either assume cooperation on the part of the agents or assume that the agents are completely self-interested, for example, in the case of bidding and other market-based approaches. However, an interesting class of MAS is one that is fundamentally cooperative, yet open, and in which one or more of the agents may be self-interested. In such systems, there is the potential for agents to misbehave, i.e., to be

miscreants

. Detecting this is tricky and context-dependent. Even more difficult is the problem of detecting collusion between agents.

In this paper, we report on a project that is beginning to address this problem using a context-based approach. Features of the MAS’ situation are used by a subset of the agents to identify it as an instance of one or more known contexts. Knowledge the agent(s) have about those contexts can then be used to directly detect miscreant behavior or collusion or to select the appropriate technique for the context with which to do so. The work is based on context-mediated behavior (CoMB), and it develops a new form of collusion detection called society-level analysis of motives (SLAM).

This paper presents an approach for contextualizing an event-based decision support system for scheduling patient assessments in a hospital. To cope with unexpected delays, patient coordinators often pursue a worst case scenario when scheduling patient assessments, leading to an underutilization of human resources and equipment when the procedure went without complications. We present a context-based decision support system for patient planning that helps the patient coordinator with taking well-informed rescheduling decisions and anticipating changes in other patients’ schedules. The system uses information and events produced by medical equipment. As these events can be non-deterministic, we demonstrate how our domain specific context model can be used to contextualize events to enhance their quality and ascertain their meaning.

This paper presents an architecture and a prototype for querying and visualising context information. The COINS modelling approach is employed to define and record the target information and its context information of an application domain. Two querying methods are provided for retrieving the recorded information: 1) natural language query which targets at physical context information and; 2) visual query environment which aims at logical context information. A knowledge representation model, master dictionary, is employed to facilitate the analyses of natural language queries. A prototype system based on the puzzle game Sudoku is used to illustrate the visual querying method, which integrates a replaying mechanism for the recorded context information.