In Nejdl, W. & De Bra, P. (Eds.). AH 2004, 26-29 August 2004, Eindhoven, The Netherlands, LNCS 3137,
Springer-Verlag, pp. 235-244, (c) Springer-Verlag, http://www.springer.de/comp/lncs/index.html
Context-Aware Recommendations in the Mobile
Tourist Application COMPASS
Mark van Setten, Stanislav Pokraev, Johan Koolwaaij
Telematica Instituut, P.O. Box 589, 7500 AN, Enschede, The Netherlands
{Mark.vanSetten, Stanislav.Pokraev, Johan.Koolwaaij}@telin.nl
Abstract. This paper describes the context-aware mobile tourist application
COMPASS that adapts its services to the user’s needs based on both the user’s
interests and his current context. In order to provide context-aware
recommendations, a recommender system has been integrated with a context-
aware application platform. We describe how this integration has been
accomplished and how users feel about such an adaptive tourist application.
1 Introduction
With several mobile technologies like mobile data networks (GPRS and UMTS),
positioning systems (GPS), mobile phones and personal digital assistants (PDAs)
getting more mature, it becomes possible to offer online services to people whenever
and wherever they are. Such online services are especially useful for people in places
they have never been to before. Apart from business travellers and truck drivers, a
large group of such people consists of tourists. Often, tourists do not know their way,
nor which restaurants, museums, shops, public services, etcetera are available to
them. The number of potential places to visit can be quite overwhelming, especially in
touristic regions. Adaptive systems can help a tourist to find places matching his
interests and his current situation.
In this paper, we consider two such adaptive systems: recommender systems and
context-aware systems. We describe their integration in a mobile tourist application
and how users feel about adaptive systems providing context-aware
recommendations. We start by introducing context-awareness and recommender
systems and how these two types of adaptive systems enhance each other (section 2).
This is followed by an overview of our mobile tourist application COMPASS (section
3). Section 4 describes the architecture of COMPASS and the underlying platform
focussing on the integration of the recommender system and context-awareness
system. Section 5 discusses the results of a survey on the usefulness of this
combination according to possible users. Section 6 ends this paper with conclusions
on context-aware recommendations.
2 Mark van Setten, Stanislav Pokraev, Johan Koolwaaij
2 Context-Aware Recommendations
Context is any information that can be used to characterize the situation of an entity.
An entity is any person, place or object that is considered relevant to the interaction
between a user and an application, including the user and application themselves [4].
Examples of contextual information are location, time, proximity, user status and
network capabilities. A general definition of context-aware systems is given in [4]: “A
system is context-aware if it uses context to provide relevant information and/or
services to the user, where relevancy depends on the user’s task.”
The key goal of context-aware systems is to provide a user with relevant
information and/or services based on his current context. This goal matches with the
goal of recommender systems. Resnick and Varian [9] define recommender systems
as systems that use opinions of a community of users to help individuals in that
community more effectively identify content of interest from a potentially
overwhelming set of choices. However, recommender systems do not only have to
incorporate the opinions of other users, but may also use other methods, such as
content-based reasoning. For this reason, we define recommender systems as systems
capable of helping people to quickly and easily find their way through large amounts
of information by determining what is of interest to a user [13]. Both context-aware
systems and recommender systems are used to provide users with relevant
information and/or services; the first based on the user’s context; the second based on
the user’s interests. Therefore, the next logical step is to combine these two systems.
Context and interests can be used as hard or soft criteria in the selection of relevant
services. Hard criteria are used to limit the set of available information and/or
services; those services that do not match a hard criterion are discarded from the
result set. Soft criteria are used to order the set of selected services or to present a
relevance score to the user for each selected service. For example, location, by far the
most exploited context factor, can be used to select only the services within a certain
distance from the user (hard criterion); location can also be used to decrease the
predicted relevance of a service the further away that service is located from the user
(soft criterion). In recommender systems, the interests of a user are mostly used as
soft criteria where the predicted level of interest is presented as a score, using for
example a number of stars. However, interests can also be used as hard criteria by
only selecting services that match the users’ interests. In our application COMPASS,
location is used as a hard criterion to select relevant services that are close to the user;
the predicted interest of the user is used as a soft criterion, just like some other
contextual factors (see section 4.4).
3 The COMPASS Application
COMPASS is an acronym for COntext-aware Mobile Personal ASSistant and is an
application that serves a tourist with information and services (ranging from buildings
to buddies) needed in his specific context that are interesting to him given his goal for
that moment. For example, a tourist expressing an interest in history and architecture
is served with information about nearby monuments built before 1890. A tourist
Context-Aware Recommendations in the Mobile Tourist Application COMPASS 3
expressing the wish to find a place for the night gets a list of hotels and campsites in
and around town that match his preferences for accommodations.
Fig. 1. Screenshots of the COMPASS application: objects near the user on the map, interacting
with services offered by objects and a list of objects near the user with relevance scores.
After start-up, COMPASS shows the user a map of his current location. The location
is either obtained from the mobile network or from other devices such as GPS
receivers. Depending on the user’s profile and goal, a selection of nearby buildings,
buddies and other objects is shown on the map and in a list. The map and the objects
shown are updated when the user moves or his profile or goal changes. Other context
changes might also force the map to change. For example, an increase in the user’s
speed by starting to drive in a car causes the map to zoom out automatically as the
user’s notion of nearness can be defined by what he can reach in a certain amount of
time; zooming out also avoids overly frequent updates of the map. Clicking on objects
on the map usually means interacting with services provided by that object (see
middle image of figure 1), e.g. calling a buddy, reserving a table at a restaurant, or
booking tickets for a show.
The application is built upon the WASP platform (see section 4.2) that provides
generic supporting services, such as a context manager and service registry. The
platform is open, which means that third parties can easily integrate their information
and services with the platform, and transparently be found and used by the population
of COMPASS users. For example, an organization that owns a collection of digitized
old postcards wrapped its database with postcards as an internet-accessible web
service, published the web service in the public service registry of the platform and
related the web service’s interface to the registry’s ontology. The net effect is that all
COMPASS users with an interest in such postcards are now able to view postcards
depicting objects near their location instantaneously. Depending on the visualisation,
they see a map of their environment with icons indicating the location depicted on the
old postcards (see the left image in figure 1) or a thumbnail list of the postcards.
Clicking on an icon displays the postcard, the date of the picture and a short
4 Mark van Setten, Stanislav Pokraev, Johan Koolwaaij
description. This way, it is quite easy to recall the atmosphere of early times while
walking through a street or neighbourhood.
The COMPASS application accomplishes this functionality by querying the
service registry for search services that are bound to deliver objects related to the
user’s context. The underlying platform retrieves services matching the hard criteria
of the user’s context and goal. For example, for someone located in Enschede and
looking for sightseeing attractions it delivers search services for museums, landmarks,
architectural buildings, etc. Next, the relevant search services are queried to retrieve
the objects matching the context’s hard criteria, e.g. to be within a certain radius from
the location of the user. The retrieved objects are then sent to the recommendation
engine which scores each object based on the soft criteria, such as the user’s interests
and contextual factors like the last time an object was visited. The retrieved objects
and scores are then displayed on the map and in the list of objects (see figure 1).
The open platform underlying the COMPASS application makes it easily
applicable in other domains as well. Examples include a mobile assistant that enables
you to remotely request a taxi to your current location or an assistant helping you to
find a house for sale in the neighbourhood that fits the wishes and desires of you and
your family. Selecting a particular house interacts with the estate agent’s service for
more information or to arrange a visit.
3.1 Related work
There are related research projects in the tourist domain using adaptive systems. The
Intrigue system [2] is an interactive agenda offered by a tourist information server that
assists the user in creating a personalised tour along tourist attractions. This research
focuses on planning and scheduling a personalised tour taking into account the
location of each tourist attraction and the interests’ of the user. Console et al. [3]
created a prototype system called MastroCARonte, which provides personalised
services that adapt to the user and his context onboard cars. This research focuses on
the effects of having such adaptive systems onboard cars.
The research focus of the COMPASS system is on the open platform, which allows
easy creation of context-aware personalised applications and the services that are part
of such a platform, including a service registry, a context manager and a
recommendation engine. The next section discusses this open platform with a focus
on the context manager and recommendation engine.
4 System Architecture
In the discussion of the architecture of the open WASP platform underlying
COMPASS and the architecture of the COMPASS application itself, we focus on the
retrieval of services taking into account context and user’s interests, as the topic of
this paper the integration of context-awareness and recommendations.
The overall architecture of the WASP platform and the COMPASS application is
shown in figure 2.
Context-Aware Recommendations in the Mobile Tourist Application COMPASS 5
Context
manager
Service
registry
Matchmaker
3G Network
Services
Recommendation
Service
Users
User profiler
Recommendation
engine
POI Retriever
Interaction
manager
WASP Platform
COMPASS
•Identification
•Charging
•Call setup
•Messaging
•etc.
Context
Services
•Location
•Time
•Weather
•Shopping list
•Agenda
•etc.
Business
Services
•Museums
•Restaurants
•Shops
•Cinemas
•etc.
Map
service
Request
dispatcher
Notification
manager
Fig. 2. The overall architecture of the WASP platform and the COMPASS application.
Four main groups can be identified in the architecture: third party services, the WASP
platform, the COMPASS application and the recommendation service.
4.1 Third party services
The 3G (GPRS, UMTS) network services provide network access capabilities, such as
user identification, call setup, messaging, charging, etc. These network capabilities
are accessible via web services interfaces and offered by mobile network operators.
The context services provide information about the context of a user, e.g. the user
status (free or busy), his location, etc. Some of this information is obtained from the
3G network via web services. This group includes both services that provide
information about the user such as his shopping list or his schedule, as well as
services that are independent from the user but which might be relevant when
selecting services, e.g. weather or traffic information services.
Business services are those services that offer information and services for an
application build on the platform. In the COMPASS application these are businesses
that offer so-called points of interest (POI): museums and their catalogues,
monumental buildings and historical information associated with them, restaurants
and their menus, shops and their current promotions, hotels with reservation services,
digitized old postcards, etc.
