L. Ardissono, P. Brna, and A. Mitrovic (Eds.): UM 2005, LNAI 3538, pp. 448 – 452, 2005. © Springer-Verlag Berlin Heidelberg 2005

Towards User Modeling Meta-ontology

Michael Yudelson1, Tatiana Gavrilova2, and Peter Brusilovsky1

1 School of Information Sciences, University of Pittsburgh, Pittsburgh PA 15260, USA

{mvy3, peterb}@pitt.edu 2 Department of Intelligent Computer Technologies,

St.-Petersburg State Technical University, Polytechnicheskaya St. 29, 195251 St.-Petersburg, Russia

gavr@csa.ru

Abstract. The paper proposes meta-ontology of the user modeling field. Ontology is meant to structure the state-of-the-art in the field and serve as a central reference point and as a tool to index systems, papers and learning media. Such ontology is beneficial for both the user modeling research community and the students as it creates a shared conceptualization of the known approaches to building user models and their implementations.

1 Introduction

User modeling (UM) is a heterogeneous field. The terminology is still not standardized. A lot of terms have multiple synonyms (e.g. behavioral user model, feature-based user model, individual user model). Terms are often fuzzy. There exist a lot of approaches to UM but a common schema that would attempt to classify them has not been proposed yet. Such lack of structure makes it harder to conduct novel research or implement known approaches in the area of UM.

This paper presents our attempt to develop a meta-ontology (a top level classify- cation) of UM field. The role of such User Modeling Meta-Ontology (UMMO) is manifold. First, it can serve as a central reference point, just like ACM computing classifycation system [1]. It is an important uniform framework to structure UM field in general. UMMO can help to present in a condensed visual form the current state of the art.

Second, such ontology will serve as a structuring basis for UM resources. It will help to collate a great number of UM approaches, papers, and systems by indexing them with the terms from ontology to be able to compare them and/or relate them to each other. The presence of a well-developed UMMO will help index coherently various information resources such as projects, papers as well as enable semi-automatic indexing.

Third, UMMO may serve as an educational tool. It will help to convey the state-of-the-art in UM field and to eliminate the synonymy problem. Finally, we hope to use UMMO as a basis to develop an adaptive educational hypermedia system for UM. Since modern adaptive hypermedia systems rely on advanced concept indexing, ontology development should be the first step toward such a system.

Towards User Modeling Meta-ontology 449

2 Known Approaches

Previously there have been several attempts to create some kind of shared conceptualization of user models. In [8] authors define user model as a set of hierarchies of personal and demographic characteristics, domain [in]dependent data, context, and low-level sensor data. Authors of [3] (mainly Peter Dolog) yet have another view of user model externalization. They view user modeling from the point of standards (such as IEEE PAPI, IMS LIP, or ISO). Although presence of standards constrains the flexibility of the user model description, but in return a more general solution is achieved, that can be applied to virtually any domain user models can be used it. The approach of the authors of the Onto-logging project [13] focuses on collaborative conceptualization in developing domain ontologies.

The first two of the mentioned above approaches make an emphasis on decomposing the user model itself. They do not touch upon the methodology surrounding user modeling, let alone classification of such methods and systems that implement them. The approach of the authors of [13] comes closest to our intentions, yet lacks some specification and structure (namely it is not an ontology). We propose an attempt to create a meta-ontology that would help to present a more general view of the user modeling field.

3 Ontology Development Process

Ontology being a useful structuring tool provides an organizing axis to help mentally mark the vision in the information hyper-space of the domain knowledge. Fig. 1 presents our view on the mainstream state-of-the-art categorization in ontological engineering [15], [8].

Fig. 1. Ontology classification (mixed ontology)

Creating ontology is a procedure that cannot be fully automated since ontology development is rather creative than formal process. Major works in this field deal with syntax problems, not semantics. Until now, only few effective domain-independent methodological approaches have been reported [12], [15], and [4]. However, in practice each development team usually follows their own set of principles, design criteria, and steps in the ontology development process.

450 M. Yudelson, T. Gavrilova, and P. Brusilovsky

Generalizing our experience in developing different teaching ontologies for e-learning in the field of artificial intelligence and neurolinguistics [4], [6], [7] we propose a four-step algorithm that may be helpful for visual ontology design.

1. Glossary development – selecting and verbalizing the essential domain concepts. 2. Laddering – defining the main levels of abstraction. It is also important for the next

stages of the design to elucidate the type of ontology according to Fig. 1 classification.

3. Disintegration/Categorization –breaking high level concepts into a set of detailed when needed (top-down strategy) and associating similar concepts to generalize meta-concepts (bottom-up strategy.

4. Refinement – updating the visual structure by excluding the excessiveness, synonymy, and contradictions.

The main goal of the algorithm above is to create a visually appealing ontology that means that ontology developer should observe conceptual balance (‘harmony’) and clarity of the ontology. A well-balanced ontological hierarchy equals a strong and comprehensible representation of the domain knowledge. Here are tips on how to achieve ‘harmony’. First, sibling concepts should be linked to the parent concept by one type of relationship. Second, the depth of the branches should be more or less equal (±2 nodes). Third, the general outlay should be symmetrical. Fourth, cross-links should be avoided as much as possible.

Ontology clarity can be achieved by optimizing the number of concepts and types of the links between them. Minimizing the number of concepts is the best tip according to Ockham’s razor principle. The maximal number of branches and the number of levels should also follow 7±2 rule by Miller. The type of relationship should be clear and obvious if the name of the relationship is omitted.

4 UMMO Development

The current version of UMMO was developed in part by extracting information about user modeling domain from various sources ([2], [10], [11], and [14]) and in part by eliciting knowledge of experts (auto-elicitation of co-authors’ knowledge so far).

Fig. 2. User Modeling Meta-Ontology upper tier

Towards User Modeling Meta-ontology 451

The process of UMMO development was guided by the aforementioned algorithm. At the glossary development stage performed both semi-automatically (keyword extraction) and manually a set of roughly 150 terms was extracted up to this moment. The laddering, generalization, and refinement stages have gone through 10 extensive iterations. A special attention was paid to generalization, since some of the concepts in the UM field have multiple terms associated with them. E.g. behavioral user model, feature-based user model, and individual user model are all synonyms. Monosemic terms were grouped in clusters, the dominated term was chosen at the authors’ discretion.

UMMO is a mixed ontology: the top layer is a partonomy, the lower layers are taxonomies. The most stable upper tier of the ontology is shown in Fig. 2. Fig. 3 shows the top part of the UMMO that contains 37 concepts (only a subset of the full ontology due to space limitation). Along with concepts Fig. 3 shows ‘markup’ nodes that depict systems/approaches (10 nodes) and people (12 nodes) that fit the concepts (systems/approaches and people are displayed as rectangles and rounded rectangles respectively).

Fig. 3. User Modeling Meta-Ontology (UMMO) upper tier and one of the branches

5 Conclusions

The challenge of such meta-ontology construction is not to identify the lower tier concepts that correspond to the individual approaches, but to work out and verbalize the upper tier concepts that would help generalize about UM methodology. UMMO is an attempt to externalize the current approaches, techniques, and tools.

452 M. Yudelson, T. Gavrilova, and P. Brusilovsky

The development of UMMO is a part of wider research aimed at development of user model centered learning portal. The UMMO development is a work in progress. Comments and suggestions would be highly appreciated at www.pitt.edu/~mvy3/ ummo_index.htm and/or at http://ummo.blogspot.com.

References

1. ACM Computing Classification System (n.d.). Retrieved November 16, 2004, from http://www.acm.org/class

2. Brusilovsky, P.: Methods and techniques of adaptive hypermedia. In: P. Brusilovsky, A. Kobsa and J. Vassileva (eds.): Adaptive Hypertext and Hypermedia. Dordrecht: Kluwer Academic Publishers (1998) 1-43

3. Dolog, P., Nejdl, W.: Challenges and Benefits of the Semantic Web for User Modelling. In Proc. of AH2003 workshop at 12th World Wide Web Conference, Budapest, Hungary (2003)

4. Fensel, D.: Ontologies: Silver Bullet for Knowledge Management and Electronic Commerce. Springer-Verlag, Berlin (2001)

5. Gavrilova, T.: Teaching via Using Ontological Engineering. In Proceedings of XI Int. Conf. “Powerful ICT for Teaching and Learning” PEG-2003, St. Petersburg (2003) 23-26

6. Gavrilova, T.A., Voinov, A.: Visualized Conceptual Structuring for Heterogeneous Knowledge Acquisition. In Proceedings of International Conference on Educational Multimedia and Hypermedia, EDMEDIA'96, MIT, Boston, USA (1996) 258-264

7. Gavrilova, T., Voinov, A.: Work in Progress: Visual Specification of Knowledge Bases. In del Pobil, A.P., Mira, J., Ali, M. (Eds), Lecture Notes in Artificial Intelligence 1416 “Tasks and Methods in Applied Artificial Intelligence”. Springer (1998) 717-726

8. Gómez-Pérez, A., Fernández-López, M., Corcho, O. Ontological Engineering with examples from the areas of Knowledge Management, e-Commerce and the Semantic Web. Springer (2004)

9. Heckmann, D.: (2003. December 22). User model ontology. (Online), retrieved on February 2, 2005. http://www.u2m.org/2003/02/UserModelOntology.daml

10. Jameson, A.: Numerical Uncertainty Management in User and Student Modeling: An Overview of Systems and Issues. User Modeling and User-Adapted Interaction, 5, (1996) 193-251

11. Kobsa, A.: User Modeling: Recent Work, Prospects and Hazards. In M. Schneider-Hufschmidt, T. Kühme, and U. Malinowski (Eds.). Adaptive User Interfaces: Principles and Practice. Amsterdam, Netherlands: North Holland Elsevier (1993)

12. Mizogushi, R. and Bourdeau J.: Using Ontological Engineering to Overcome Common AI-ED Problems. International Journal of Artificial Intelligence in Education, 11, (2000) 1-12

13. Razmerita, L., Angehrn, A., and Maedche, A.: Ontology based user modeling for Knowledge Management Systems. In Proceedings of the 9th International Conference on User Modeling, Pittsburgh, USA, 2003, Springer-Verlag (2003) 213-217

14. Rich, E.: User Modeling via Stereotypes. In M. Maybury & W. Wahlster (Eds.). Readings in Intelligent User Interfaces. Morgan Kaufmann Press, CA: San Francisco (1998) 329-342

15. Tu, S., Eriksson, H., Gennari, J., Shahar, Y. & Musen M.: Ontology-Based Configuration of Problem-Solving Methods and Generation of Knowledge-Acquisition Tools. In Artificial Intelligence in Medicine, N7 (1995) 257-289

16. Uschold, M., Jasper, R.: A Framework for Understanding and Classifying Ontology Applications. In Proceedings of the IJCAI-99 Workshop on Ontologies and Problem-Solving Methods (KRR5), Stockholm, Sweden (1999)