Judy Kay - Ontologies for personalisation projects
 Home    Bio   Research   Teaching
Ontologies for personalisation
A software representation of an extensive computer science ontology may be used in a variety of roles in a teaching system. It enables a system to infer from a minimal knowledge base about a student to a quite extensive model of their knowledge. Teaching goals can then be inferred by studying this model and expanding concepts the student is modelled as knowing working with those that are closely related. It is also possible for a learning goal to be established, with the ontology identifying aspects that need to be taught as a foundation for th is goal; for example, if they are prerequisites. The size and scope of such an ontology means that it becomes necessary to effectively query the information it represents. This also allows a model within the scope of the ontology to be constructed efficiently using minimal knowledge. Models can be compared, looking for similarities, and a result given that reflects their relative structure and position in the ontology. This result can then be used as additional information to learn from the models.
Automatic Construction of Learning Ontologies
Trent Apted and Judy Kay
Rather than creating the ontology by hand, we have built MECUREO, a tool for automatically constructing an ontology. It uses an existing, reliable resource to ensure appropriate coverage of computer science terms. This saves a considerable amount of effort and helps minimise errors while maximising breadth of coverage. The generated ontology may also serve as a basis for refinement by hand. For example, it is often easier to categorise an unknown relationship (that is, a relationship known to exist for which there is not enough information to categorise it) by examining the automatic construction than it is to identify a relationship independently.
Links: User model grower; Examples of queries.

Publications:
 

Apted, T and J Kay, (2002) Generating and Comparing Models within an Ontology, Kay, J and J Thom, (eds) Proceedings ADCS2002, Australian Document Computing Symposium, 16 December 2002, to appear. online proceedings

 

Apted, T and J Kay, (2002) Automatic Construction of Learning Ontologies, Aroyo, L and Dicheva, D, (eds) Proceedings ICCE Workshop on Concepts and Ontologies in Web-based Educational Systems, ICCE 2002, International Conference on Computers in Education, CS-Report 02-15 Technische Universiteit Eindhoven, on-line proceedings 55-62.

Verified Concept Mapping for Eliciting Conceptual Understanding
Laurent Cimolino, Judy Kay and Amanda Miller
Concept mapping is a valuable technique for education evaluation. Concepts maps have become a common tool for externalising learner conceptions of a domain. There are many available tools for learners to draw concept maps. Concept mapping has strong foundations in theories of learning and in empirical studies of brain activity. The approach has many potential roles in education. We are concerned with the use of concept maps as a mechanism for determining the way that a learner conceptualises a domain. From this, we aim to build accurate and detailed learner models of the learner's conceptual knowledge of a domain. Since our research focuses on scrutable learner modelling and strong learner control, the concept map is a natural tool for eliciting learner models. In line with our philosophy of learner control, our current work explores approaches to building verified concept maps for the purpose of modelling the student's knowledge. We consider it critical to verify concept maps before using them as a basis for reasoning about the student's knowledge. This is because it is very easy for a student to accidentally link the wrong concept or omit a concept or a link from their concept map. Moreover, revision of concept maps is a normal and important part of the concept mapping process. The vcm tool supports teachers in defining effective diagnostic concept mapping tasks so that students can provide a detailed, verified model of their understanding or personal ontology for a domain.
Source code is available. If you take a copy and want to use it, please let me know what you are doing. If you make some enhancements, I would be keen to know about that.
Acknowledgement: This software has been built from an excellent foundation provided by JGraph written by Gaudenz Alder.
Publications:
 

Cimolino, L and J Kay, (2002) Verified Concept Mapping for Eliciting Conceptual Understanding, Aroyo, L and Dicheva, D, (eds) Proceedings ICCE Workshop on Concepts and Ontologies in Web-based Educational Systems, ICCE 2002, International Conference on Computers in Education, CS-Report 02-15 Technische Universiteit Eindhoven, on-line proceedings 9-14.

Current projects available:
An essential next step for the verified concept mapper is to evaluate its effectiveness. This requires construction of concept mapping tasks followed by evaluation of them and the vcm tool in experiments where learners use vcm to show their understanding of an area.
A really interesting next stage for the Mercurio work is to explore the use of WordNet as a basis for evaluating the ontologies constructed from dictionaries. This is part of the very difficult problem of evaluating ontologies that have been automatically constructed from existing document sources. There are several other projects in this broad area including, for example, evaluating one of the ontologies in a real application, such as supporting a teaching system for C or user interface design.
There is a natural link between these two projects. Already, we have performed some evaluations of Mercurio by asking users to construct concept maps on paper and then comparing these with Mercurio's ontologies as a means of evaluating them. A natural next step is to link vcm's maps with evaluation of Mercurio ontologies.
Another natural link is in the other direction, where people create free form concept maps using vcm and the relevant point queries on a Mercurio ontology are used to suggest additional links. This would be an invaluable tool for people constructing ontologies by hand, for a special purpose but with assistance from the Mercurio ontology.
I am really keen to see the vcm work linked to formal work on reasoning about diagrams. At present the relationships and reasoning in vcm are quite intuitive but ad-hoc. It would useful to explore the work of Ken Forbus's group as a formal basis for reasoning about concept maps. See:

Kenneth D. Forbus and Ronald W. Ferguson and Jeffery M. Usher, 2001, Towards a computational model of sketching, Intelligent User Interfaces, 77-83. citeseer.nj.nec.com/422041.html.

Brian Falkenhainer and Kenneth D. Forbus and Dedre Gentner, 1989, The Structure-Mapping Engine: Algorithm and Examples, Artificial Intelligence, 41(1), 1-63. citeseer.nj.nec.com/falkenhainer89structuremapping.html


This area links with SIT-CRC projects on modelling individuals and communities. There are several important projects you should come and discuss.
Top of Page