ADAPTING THE SUPPLY OF CULTURAL EVENTS TO DIFFERENT KINDS OF CONSTRAINTS J´erˆome Fuselier, Jean-Charles Marty, Laurence Vignollet Laboratoire SYSt`emes COMmunicants de l’Universit´e de Savoie [email protected] [email protected] [email protected]

Abstract : R (Electronic Schoolbag) project was born in 1999, from The ”Cartable Electronique ” a collaboration between the ”Syst`emes Communicants” team from University of Savoie, the Savoie General Council and the State Education Services. The electronic schoolbag is the technical means by which pupils organize their nomadism between home and school, and keep ”in hand” useful tools or data in every environment they access. This metaphor directly addresses student mobility and the personalization of the content, services and interface given. As usual in most adaptive systems, the user characteristics are central to this personalization. However, it seems very interesting to be able to consider other constraints, like for instance the graphical charter, or those linked to the information type (for a film description, video extracts should be highlighted), or linked to the device. New problems thus appear. How to select the right pieces of information to construct the bestsuited page for the user ? How to represent the different constraints ? How to manage the different conflicts due to constraints coming from different sources? How to take into account layout constraints (on screen areas) in adaptive hypermedia in a general way ? These different items are addressed in this paper, through an integrative architecture and the definition of small languages based on XML. Keywords : Adaptive Hypermedia, automatic generation, document model, XML, electronic schoolbag, user model 1

Introduction

R project was born in 1999, from a collaboration The ”Cartable Electronique ” between the ”Syst`emes Communicants” team from University of Savoie, the Savoie General Council and the State Education Services. The electronic schoolbag is the technical means by which pupils organize their nomadism between home and school, and keep ”in hand” useful tools or data in every environment they access. This metaphor directly addresses student mobility and the personalization of the content, services and interface given. From the beginning, efforts were put into the development of an online collaborative environment, giving access to applications and services and into the development of these applications and services themselves. This second axis requires the implication of the traditional educative partners, also in culture or sport areas.

Concerning the cultural aspects, we noticed that, even when cultural broadcast systems exist, they do not adapt to the user and usually provide static information. In the Joensuu, Finland, 2002

R context of the ”Cartable Electronique ” project, we have developed an adaptive hypermedia [Brusilovsky, 2001] allowing cultural partners to put together the cultural events [Marty and Vignollet, 2001]. The user can find in his electronic schoolbag, the targeted and adapted cultural palette : the service builds dynamically an adapted event presentation using the cultural fragments of information given by the partners, based on the user characteristics. As usual in most adaptive systems, this first version takes only into account the user characteristics. However, it seems very interesting to be able to consider other constraints, like for instance the graphical charter (e.g. the charter sets the size and localization of the partner’s logo), or those linked to the the device.

New problems thus appear. How to select the right pieces of information to construct the best-suited page for the user ? How to represent the different constraints ? How to manage the different conflicts due to constraints coming from different sources ? How to take into account layout constraints (on screen areas) in adaptive hypermedia in a general way? In this paper, we propose an architecture to take into account these problems. We will see that the constraints doesn’t have the same role in the adaptation and that we will have to classify them. We also propose a new language for the interface description, taking into account new trends like XIML [Puerta and J. Eisenstein, 2002]. 2

Fragments

The domain data are constituted from pieces of information (that we call ”fragments”) given by the cultural partners. These fragments can be for instance a text description of an event, pictures of this event, interviews related to this event, etc... In order to select the appropriate fragments (the best suited fragments according to some constraints), we need to have meta data on these fragments. For instance, a fragment is dedicated to young children ; or a fragment is a textual abstract of an event, . . . The use of fragments, as it can be found in other systems [Ardissono et al., 2001], offers a large number of adaptation possibilities. Let’s consider for instance the case of a young child and of his teacher consulting the same cultural event. As our system is an adaptive one, the generated page won’t be the same in the two cases: the selected items of information are different with regard to the meta data provided with the fragments. Some are relevant for the teacher (the compositor biography) while others are relevant for the child (an attractive picture of the event). The meta data on the fragments are expected by the system. When the partners enter the fragments, the system asks for additional information (meta data). 3

User Model

In our system, we consider a long term dynamic user model. Long term, because we need to know the evolution of the user’s cultural preferences as well as his presentation preferences. Dynamic, because it is appropriate to adapt the User Model according to the user’s actions in the system (in particular by examining the ’surprise factor’, i.e. the factor that indicates how far the user’s actions are from what is expected by the system). The User Model contains a set of triplets composed of the name of the attribute, the Joensuu, Finland, 2002

value of this attribute and a factor of uncertainty about this value. [Fuselier et al., 2002] 4

Constraint Classification

The information to be presented has to be adapted. Our approach is to use constraints on the information to be presented to guide this adaptation process [Boukottaya02 et al., 2002]. These constraints could be induced by the user model but also by cultural partners or by the device. For us, there are two main kinds of constraints to take into account: – the contents constraints ; – what is the information to be presented? – the presentation constraints ; – how do we present it? The contents constraints deal with the selection of the fragments. Constraints from the different sources listed above facilitate the selection of fragments : For instance, a simple case is the selection of the fragments according to the user’s cultural preferences (theater, cinema) or according to the user’s presentation preferences (prefers a synthetic text rather than a detailed one). The stereotype constraints can also guide this part of the selection process: the system will exclude the textual fragments for a young child or it will include more information for a teacher. We can also mention that some of the content constraints come from the partners : it can be important for a cultural partner to give rules about the minimal contents of an event description (at least a title, a short abstract and a picture). Device constraints can help in deciding which pieces of information to select from the capabilities of the device (e.g. images can’t be displayed on this device) and we must think of strategies to replace information that can’t be displayed, as far as it is possible. The presentation constraints deal with the graphical placement of the various objects on the page (layout) and with the appearance of these objects (style). The partner’s graphical charter is an example of a set of constraints that modify the layout and the style of the page. We can also imagine constraints linked to the user preferences (size of icons) that influence the style of objects on the page or that influence the layout of objects (the user prefers to always have an abstract at the top of the page between two images). Concerning the device constraints, the number of rows and columns modifies the layout of the page. Some other constraints such as font color or font family can modify the style of the information. We would like to stress the fact that we want to separate clearly the layout and the style. In this way, it will become possible to clearly specify ”areas” on the screen where some constraints could apply, which is not currently the case in the existing adaptive systems, to our knowledge. An area can be seen as a container. It corresponds to a location where the information can be placed during the generation process. We can have constraints on an area: this can be viewed as the association of some properties to an area, e.g. the background color or the font used for the style, the position attributes for the layout. 5

Architecture

The figure fig.1 reflects the architecture of our system. It explains how we generate a page from the different pieces of information (or ”fragments”) given by the cultural Joensuu, Finland, 2002

partners. We can see that this generation depends on different constraints coming from different sources (the user model is one of them). The ”interface description generation” box contains the representation of what should be generated. Fragments

User

Content Constraints

Fragments Selection

Device Presentation Constraints

Cultural Partner

Interface Description Generation

Interface Generation

F IG . 1 – The architecture of our system

5.1

Fragments selection

The first stage consist in determining, from available fragments and from the content constraints, a subset of relevant fragments. The selection is realized with the meta-datas associated with every fragment. 5.2

Interface description generation

In most adaptive hypermedia like AHA! [Bra and Ruiter, 2001] or Tellim [J¨ording, 1999], the interface generation is ”hard-coded” and it is rather difficult to change the appearance of a generated page according to the user. In our application, we think that the description of the interface is an important part of the system, since the generation will be done from this description and according to the target language, and to the device used by the user. We have defined a language in XML which allows the description of the structure of the page, independently from the language which will be used finally. The combination of the different constraints and of the fragments which are selected to be displayed is sufficient to generate such a description. This description contains the structure of the page, represented by screen areas and fragments, the style to be used for every zone and element, the effective information to be displayed. You can find a more precise example of the description language in Fuselier [Fuselier, 2002]. Joensuu, Finland, 2002

5.3

Interface generation

The relative independence of the interface description language allows us to generate an interface in a large number of programming languages. The only needed development is a parser which performs the transformation from the source language, in XML, to the target language, like HTML or Java for instance. Such a component has been realized thanks to the XSLT transformation language. 6

Conclusion

Our work aims at improving the man machine interfaces. We want to adapt the cultural information supply to different users. The specification and the implementation of a first R has already been completed. version of this service of the ”Cartable Electronique ” We have developed a system which allows cultural partners to enter information about cultural events. It is an adaptive system, since the cultural events are not presented in the same way to the different kinds of users. Now, we would like to improve our system in order to create an adaptive service which is not dependant on the application domain. The user model has to be expanded, and the dynamic generation of the document for a given user must be precisely described. This paper is more concerned with this last topic. We believe that approaching the problem from the constraints point of view can make our system more general and more powerful. The possibility of using constraints on areas opens up new possibilities for adaptation. Currently, we are in the design stage of the new version. We are able to represent the layout constraints by using a general formalism that we have specified. These constraints can be taken into account in order to generate an interface description that is quite independent from the target language (e.g. HTML, JAVA). We have also developed the translator allowing to generate HTML pages from the interface description. A described interface could then be integrated into any application. Although we can express constraints for the partner on the final presentation of the information, the next step is to refine the description language of these constraints to make it more general and allow to take constraints from different sources into consideration. At that stage, we will address the problem of the constraints conflict resolution between the different sources. R´ef´erences [Ardissono et al., 2001] Ardissono, L., Console, L., and Torre, I. (2001). An adaptive system for the personalized access to news. AI Communications, 14(3):129–147. [Boukottaya02 et al., 2002] Boukottaya02, A., Vanoirbeek, C., Ngoc, A. N., Rekik, Y., and Zeramdini, K. (2002). A contract based model for creating structured virtual documents : key issues for reusability and collaboration. In DVP’02. [Bra and Ruiter, 2001] Bra, P. D. and Ruiter, J. (2001). Aha! adaptive hypermedia for all. In Proceedings of the WebNet Conference, pages 262–268. [Brusilovsky, 2001] Brusilovsky, P. (2001). Adaptive hypermedia. User Modeling and User Adapted Interaction, pages 87–110. Ten Year Anniversary Issue. Joensuu, Finland, 2002

[Fuselier, 2002] Fuselier, J. (2002). Composition de fragments d’informations en vue d’une pr´esentation adapt´ee aux utilisateurs. Master’s thesis, Universit´e de Savoie. [Fuselier et al., 2002] Fuselier, J., Marty, J., and Vignollet, L. (2002). Une g´en´eration automatique de documents virtuels personnalis´es guid´ee par des contraintes. In DVP’02. [J¨ording, 1999] J¨ording, T. (1999). A temporary user modeling approach for adaptive shopping on the Web. In 2nd workshop on adaptive systems and user modeling on the WWW. [Marty and Vignollet, 2001] Marty, J. and Vignollet, L. (2001). un service culturel adaptatif int´egr´e au portail e´ ducatif. In Nimes’TIC. [Puerta and J. Eisenstein, 2002] Puerta, A. and J. Eisenstein, X. (2002). Ximl : A common representation or interaction data. In IUI2002. Acknowledgments: We would particularly like to express thanks to Christian Martel, manager of the ”Cartable Electronic” project and to the engineers of the TIC group of the Savoie General Council. Thanks also to the cultural partners of the project, particularly the ADMS (Association pour la Danse et la Musique en Savoie), for their involvment in the design of the cultural service.

Joensuu, Finland, 2002