Eleni Christopoulou
University of Patras & Ionian University, Greece
aBstract
This chapter presents how the use of context can support user interaction in mobile applications.
It argues that context in mobile applications can be used not only for locating users and provid-ing them with suitable information, but also for supporting the system’s selection of appropriate interaction techniques and providing users with a tool necessary for composing and creating their own mobile applications. Thus, the target of this chapter is to demonstrate that the use of context in mobile applications is a necessity. It will focus on the current trend of modeling devices, services and context in a formal way, like ontologies, and will present an ontology-based context model.
IntroductIon
The future of computer science was marked by Weiser’s vision (Weiser, 1991), who introduced the term ubiquitous computing (ubicomp) by defining a technology that can be seamlessly integrated into the everyday environment and aid people in their
everyday activities. A few years later, the Euro-pean Union, aiming to promote “human-centered computing,” presented the concept of ambient intelligence (AmI) (ISTAG, 2001), which involves a seamless environment of computing, advanced networking technology and specific interfaces.
So, technology becomes embedded in everyday objects such as furniture, clothes, vehicles, roads, and smart materials, providing people with the tools and processes that are necessary in order to achieve a more relaxing interaction with their environment.
Several industry leaders, like Philips and Microsoft, have turned to the design of ubicomp applications with a focus on smart home applica-tions. However, people nowadays are constantly on the move, travel a lot, and choose to live in remote or mobile environments. In the near future, each person will be “continually interacting with hundreds of nearby wirelessly connected comput-ers” (Weiser, 1993). Therefore, the need for mobile applications is now more evident than ever.
Recent years have seen a great breakthrough occur in the appearance of mobile phones. Initially they were used as simple telephone devices. Today,
mobiles have evolved into much more than that.
Although the majority of people still use mobile phones as communication devices, an increasing number of users have begun to appreciate their potential as information devices. People use their smart mobile phones to view their e-mails, watch the news, browse the Web, and so forth. Eventually, mobile phones and other mobile handheld devices became an integral part of our daily routine.
Both scientists and designers of ubicomp appli-cations have realized that the mobile phone could be considered as one of the first AmI artefacts to appear. As mobile phones are becoming more pow-erful and smarter this fact is increasingly proven true. Thus, scientists wanting to take advantage of the emerging technology have implemented a great number of mobile applications that enable human-computer interaction through the use of handheld devices like mobile phones or personal digital assistants (PDAs). Such applications in-clude visitor guides for cities and museums, car navigation systems, assistant systems for confer-ence participants, shopping assistants and even wearable applications.
A closer examination of mobile applications shows that most of them are location-aware sys-tems. Specifically, tourist guides are based on users’ location in order to supply more informa-tion on the city attracinforma-tion closer to them or the museum exhibit they are seeing. Nevertheless, recent years have seen many mobile applications trying to exploit information that characterizes the current situation of users, places and objects in order to improve the services provided. Thus, context-aware mobile applications have come to light.
Even though significant efforts have been devoted to research methods and models for capturing, representing, interpreting, and exploit-ing context information, we are still not close to enabling an implicit and intuitive awareness of context, nor efficient adaptation to behavior at the standards of human communication practice. Most of the current context-aware systems have been
built in an ad-hoc approach, deeply affected by the underlying technology infrastructure utilized to capture the context (Dey, 2001). To ease the de-velopment of context-aware ubicomp and mobile applications it is necessary to provide universal models and mechanisms to manage context.
Designing interactions among users and devices, as well as among devices themselves, is critical in mobile applications. Multiplicity of devices and services calls for systems that can provide various interaction techniques and the ability to switch to the most suitable one ac-cording to the user’s needs and desires. Context information can be a decisive factor in mobile applications in terms of selecting the appropriate interaction technique.
Another inadequacy of current mobile systems is that they are not efficiently adaptable to the user’s needs. The majority of ubicomp and mobile applications try to incorporate the users’ profile and desires into the system’s infrastructure either manually or automatically observing their habits and history. According to our perspective, the key point is to give them the ability to create their own mobile applications instead of just customizing the ones provided.
The target of this chapter is to present the use of context in context-aware ubicomp and mobile applications and to focus on the current trend of modeling devices, services and context in a for-mal way (like ontologies). Our main objective is to show that context in mobile applications can be used not only for locating users and provid-ing them with suitable information, but also for supporting the system’s selection of appropriate interaction techniques and for providing them with a tool necessary for composing and creating their own mobile applications.
In the background section, which follows, we define the term context and present how context is modeled and used in various mobile applications focusing on ontology-based context models. In the subsequent sections we present our perspective of context, an ontology-based context model for
mobile applications as well as the way in which human-computer interaction can be supported by the use of context. The Future section embraces our ideas of what the future of human-computer interaction in mobile applications can bring by taking context into account. Finally we conclude with some prominent remarks.
Background What is context
The term “context-aware” was first introduced by Schilit and Theimer (1994), who defined context as “the location and identities of nearby people and objects, and changes to those objects.” Schilit, Adams, and Want (1994) defined context as “the constantly changing execution environment” and they classified context into computing ment, user environment, and physical environ-ment. Schmidt (2000) also considered situational context, such as the location or the state of a device, and defined context as knowledge about the state of the user and device, including surroundings, situation and tasks and pointing out the fact that context is more than location.
An interesting theoretical framework has been proposed by Dix et al. (2000), regarding the notions of space and location as constituent aspects of context. According to this framework context is decomposed into four dimensions, which complement and interact with each other.
These dimensions are: system, infrastructure, domain, and physical context.
One of the most complete definitions for con-text was given by Dey and Abowd (2000); accord-ing to them context is “any information that can be used to characterize the situation of an entity.
An entity should be treated as anything relevant to the interaction between a user and an application, such as a person, a place, or an object, including the user and the application themselves.”
When studying the evolution of the term
“context” one notices that the meaning of the term has changed following the advances in con-text-aware applications and the accumulation of experience in them. Initially the term “context”
was equivalent to the location and identity of users and objects. Very soon, though, the term expanded to include a more refined view of the environment assuming either three major compo-nents; computing, user and physical environment, or four major dimensions; system, infrastructure, domain, and physical context. The term did not include the concept of interaction between a user and an application until Dey and Abowd (2000).
This definition is probably at present the most dominant one in the area.
context modeling in context-aware applications
A number of informal and formal context mod-els have been proposed in various systems; the survey of context models presented in Strang and Linnhoff-Popien (2004) classifies them by the scheme of data structures. In Partridge, Begole and Bellotti (2005) the three types of contextual models, which are evaluated, are environmental, personal, and group contextual model.
Among systems with informal context models, Context Toolkit (Dey, Salber & Abowd, 2001) represents context in the form of attribute-value tuples, and Cooltown (Kindberg et al., 2002) proposed a Web-based model for context in which each object has a corresponding Web description.
Both ER and UML models are used for the repre-sentation of formal context models in Henricksen, Indulska, and Rakotonirainy (2002).The context modeling language is used in Henricksen and Indulska (2006) in order to capture user activities, associations between users and communication channels and devices and locations of users and devices.
Truong, Abowd and Brotherton (2001) point out that the minimal set of issues required to be
addressed when designing and using applications are: who the users are, what is captured and ac-cessed, when and where it occurs, and how this is performed. Designers of mobile applications should also take these issues into account. Similar to this approach Jang, Ko and Woo (2005) pro-posed a unified model in XML that represents user-centric contextual information in terms of 5W1H (who, what, where, when, how, and why) and can enable sensor, user, and service to differently generate or exploit a defined 5W1H-semantic structure.
Given that ontologies are a promising instru-ment to specify concepts and their interrelations (Gruber, 1993; Uschold & Gruninger1996), they can provide a uniform way for specifying a con-text model’s core concepts as well as an arbitrary amount of subconcepts and facts, altogether en-abling contextual knowledge sharing and reuse in a Ubicomp system (De Bruijn, 2003). Ontologies are developed to provide a machine-processable semantics of information sources that can be com-municated between different agents (software and humans). A commonly accepted definition of the term ontology was presented by Gruber (1993) and stated that “an ontology is a formal, explicit specification of a shared conceptualization.” A
“conceptualization” refers to an abstract model of some phenomenon in the world which identi-fies the relevant concepts of that phenomenon;
“explicit” means that the type of concepts used and the constraints on their use are explicitly defined and “formal” refers to the fact that the ontology should be machine readable. Several research groups have presented ontology-based models of context and used them in ubicomp and mobile applications. We will proceed to briefly describe the most representative ones.
In the Smart Spaces framework GAIA (Ranga-nathan & Campbell, 2003) an infrastructure that supports the gathering of context information from different sensors and the delivery of appropriate context information to ubicomp applications is presented; context is represented as first-order
predicates written in DAML+OIL. The context ontology language (Strang, Linnhoff-Popien &
Frank, 2003) is based on the aspect-scale-context information model. Context information is at-tached to a particular aspect and scale and qual-ity metadata are associated with information via quality properties. This contextual knowledge is evaluated using ontology reasoners, like F-Logic and OntoBroker.
Wang, Gu, Zhang et al. (2004) created an upper ontology, the CONON context ontology, which captures general features of basic contextual enti-ties, a collection of domain specific ontologies and their features in each subdomain. An emerging and promising context modeling approach based on ontologies is the COBRA-ONT (Chen, Finin
& Joshi, 2004). The CoBrA system provides a set of OWL ontologies developed for modeling physical locations, devices, temporal concepts, privacy requirements and several other kinds of objects within ubicomp environments.
Korpipää, Häkkilä, Kela et al. (2004) present a context ontology that consists of two parts:
structures and vocabularies. Context ontology, with the enhanced vocabulary model, is utilized to offer scalable representation and easy naviga-tion of context as well as acnaviga-tion informanaviga-tion in the user interface. A rule model is also used to allow systematic management and presentation of context-action rules in the user interface. The objective of this work is to achieve personaliza-tion in mobile device applicapersonaliza-tions based on this context ontology.
Although each research group follows a dif-ferent approach for using ontologies in modeling and managing context in ubicomp and mobile applications, it has been acknowledged by the majority of researchers (Biegel & Cahill, 2004;
Dey et al., 2001; Ranganathan & Campbell, 2003) that it is a necessity to decouple the process of context acquisition and interpretation from its actual use, by introducing a consistent, reliable and secure context framework which can facilitate the development of context-aware applications.
context utilisation in mobile applications
In context-aware mobile applications location is the most commonly used variable in context recognition as it is relatively easy to detect.
Thus, a lot of location-aware mobile systems have been designed, such as shopping assistants (Bohnengerger, Jameson, Kruger et al., 2002) and guides in a city (Davies, Cheverst, Mitchell et al., 2001) or campus area (Burrell, Gay, Kubo et al., 2002). Many location-aware mobile ap-plications are used in museum environments; a survey is presented in (Raptis, Tselios & Avouris, 2005). In the survey of Chen and Kotz (2000) it is evident that most of the context-aware mobile systems are based on location, although some other variables of context like time, user’s activity and proximity to other objects or users are taken into consideration.
User activity is much more difficult to identify than location, but some aspects of this activity can be detected by placing sensors in the envi-ronment. Advanced context-aware applications using activity context information have been put into practice for a specific smart environment (Abowd, Bobick, Essa et el., 2002). The concept of activity zones (Koile, Tollmar, Demirdjian et al., 2003) focuses on location, defines regions in which similar daily human activities take place, and attempts to extract users’ activity information from their location.
Sensor data can be used to recognize the us-age situation based on illumination, temperature, noise level, and device movements, as described for mobile phones in Gellersen, Schmidt and Beigl (2002) and PDA in Hinkley, Pierce, Sinclair et al.(2000), where it is suggested that contextual information can be used for ring tone settings and screen layout adaptation. The mobile device can observe the user’s behavior and learn to adapt to a manner that is perceived to be useful at a certain location as was the case with the comMotion system (Marmasse & Schmandt, 2000).
Sadi and Maes (2005) propose a system that can make adaptive decisions based on the context of interaction in order to modulate the informa-tion presented to the user or to carry out semantic transformation on the data, like converting text to speech for an audio device. CASIS (Leong, Kobayashi, Koshizuka et al., 2005) is a natural language interface for controlling devices in intelligent environments that uses context in order to deal with ambiguity in speech recogni-tion systems. In Häkkilä and Mäntyjärvi (2005) context information is used in order to improve collaboration in mobile communication by sup-plying relevant information to the cooperating parties, one being a mobile terminal user and the other either another person, group of people, or a mobile service provider.
Perils of context-awareness
The promise and purpose of context-awareness is to allow computing systems to take action autono-mously; enable systems to sense the situation and act appropriately. Many researchers, though, are skeptical and concerned because of the problems that emerge from context-awareness.
A main issue regarding context-aware comput-ing is the fear that control may be taken away from the user (Barkhuus & Dey 2003). Experience has shown that users are still hesitant to adopt con-text-aware systems, as their proactiveness is not always desired. Another aspect of this problem is that users often have difficulties when presented with adaptive interfaces.
Apart from control issues, privacy and security issues arise. The main parameters of context are user location and activity, which users consider as part of their privacy. Users are especially reluc-tant to exploit context-aware systems, when they know that private information may be disclosed to others (Christensen et al., 2006).
Even recent research projects suffer from dif-ficulties in automated context fetching; in order to overcome this, the user is asked to provide context
0
manually. Studies have shown that users are not willing to do much in order to provide context and context that depends on manual user actions is probably unreliable (Christensen et al., 2006).
Additionally, systems that ask from users to supply context fail, as this affects the user’s experience and diminishes his benefit from the system.
Practice has shown that there is a gap between how people understand context and what systems consider as context. The environment in which people live and work is very complex; the abil-ity to recognize the context and determine the appropriate action requires considerable intel-ligence. Skeptics (Erickson, 2002) believe that a context-aware system is not possible to decide with certainty which actions the user may want to be executed; as the human context is inaccessible to sensors, we cannot model it with certainty. They, also, argue whether a context-aware system can be developed to be so robust that it will rarely fail, as ambiguous and uncertain scenarios will always occur and even for simple operations exceptions may exist. A commonly applied solution is to add more and more rules to support the decision making process; unfortunately this may lead to large and complex systems that are difficult to understand and use.
An issue that several researchers bring forward (Bardram, Hansen, Mogensen et al., 2006) is that context-aware applications are based on context information that may be imperfect. The ambigu-ity over the context soundness arises due to the speed at which the context information changes and the accuracy and reliability of the producers of the context, like sensors.
It is a challenge for context-aware systems to handle context, that may be non accurate or ambiguous, in an appropriate manner. As Moran and Dourish (2001) stated, more information is not necessarily more helpful; context information is useful only when it can be usefully interpreted.
What Is conteXt For moBIle aPPlIcatIons?
Considering the use of context in the mobile ap-plications discussed in the background section, we may conclude that, for these applications, context is almost synonymous to location and, specifically, to user location. However, context is quite more than just that. In this section, we will present our perspective on the parameters of context that are necessary for mobile applications.
In order to figure out these parameters we have
In order to figure out these parameters we have