A User Modeling Ontology for Mathematics
Related Links
Roadmap
Overview
BTW, you can use 1. item (1. in every line) for numbered lists in Trac --Christoph
1 Motivation
2 Analysing existing user modeling ontology
3 Development of a mathematical user modeling ontology
4 Representation
5 Integration into OMDoc
6 Implementation, Case Study, Evaluation
1 Motivation
Why do we need a user modeling ontology for mathematics? What is its application (where is it used)?
- so far we provide configurable service, see JOBAD, but all adaptation efforts are user-driven: e.g. users can change the notation in a document and play with the brackets
- system-driven, automatic adaptation, and memorizing configuration settings, requires to capture user settings and information: e.g. one a user has changed his notations in an active documents, the system memorizes his configuration. Moreover, his settings can be used to adapt other pages automatically (propagation of user configuration)
- good justification! --Christoph
- Maybe ground this by citing general results on user modeling. I don't know much about it, but I'd expect results like the following to be available: preferences of a single user change slowly (therefore it's good to remember them), users appreciate if the system remembers things for them. --Christoph
- Maybe extend this into communities. User is a member of a CoP <=> his preferences are the same as (or similar to) the ones of the community (no idea if that's the case, just guessing) -- use community profiles as templates for user profiles (direction =>), generate community profiles from the average user profile (direction <=) --Christoph
2 Analysing existing user modeling ontology
We describe the state-of-the-art of user modeling ontologies pointing to the domains, advantages, and drawbacks of existing user modeling ontologies. Can we reuse an approach for our mathematical domain (see next step)?
No idea if there are proper ontologies, but of course there are other formal models that just may not be named "ontologies" --Christoph
3 Development of a mathematical user modeling ontology
Reuse and extend existing ontologies for mathematics.
Analogously to the attempt of finding a versioning ontology for OMDoc, we have evaluated existing user modeling ontologies above. We need to decided whether we can directly reuse one of the existing approaches or can reuse some contepts and structures for our own ontology for mathematics. In particular, we want to model mathematical practices (and preferences), starting with notation preferences: We propose notation tags to represent explicit notation choices and different notation context to describe their use in different situations (see first ideas based on which we want to outline other practices, see e.g. the Activemath user model for an approach to represent knowledge and competencies)
- Note about these tags: always keep syntax (the tags) and semantics (the ontology, or the formal knowledge extracted from these tags) in sync. My point here is: the tags are only meaningful if backed by an ontology (i.e. background knowledge and inference rules for logical reasoning). OK, a meaningful interpretation of such a tag syntax can also be hard-coded into the application, but that's much less flexible and extensible. (I suppose that's what ActiveMath? does) --Christoph
4 Representation in RDF
Which ontology language do we choose (see state-of-the-art of ontology languages)? Why (if) do we choose to represent our ontology in RDF. What are benefits. Describe the inference mechanism and how it can forster user modeling and adaptations.
Would be interesting to find out in what languages existing user modeling ontologies are implemented. But the decentral, distributed aspect is certainly in favor of something RDF-based. Note that RDF is just the foundation; there are many different inference mechanisms on top of RDF (description logic, rules, even completely customized ones, and of course combinations). A state of the art of ontology languages is good to know but IMHO not really needed in such a paper. --Christoph
- Benefits: search, query, interlinking on the web
- Decentral, distributed approach: In the SIOC approach, each user (foaf:Person) has several user models on different sites, from which he publishes certain parts. Moreover, he creates content on these its. We can combine the profiles and his contents to infer new information.
- Central approach: We provide a central server for the management of profiles. Certain parts of these profiles are published and can be used in different sites. Sites can either import from and export to the central service, or directly integrate it (see panta rhei scenario? and panta rhei framework?)
5 Integration into OMDoc
Why do we want to integrate the RDF in OMDOc and how do we do it (RDFa ...)
This is an extension of the ESWC paper.
No more comments to add at the moment; just have to see how the ESWC project evolves. --Christoph
6 Implementation
Tools for handling RDFa (extracting from OMDoc, inference, adaptation engine)
- Extraction is covered (see ESWC paper)
- except for the extension to the user model tags, but that's similar to the existing support and therefore easy to add . --Christoph
- Inference and adaptation engine are TODOs
Potential Conference for Publication
- http://umap09.fbk.eu/ (deadline 12.01.); track: Aspects of users that are modeled and adapted to ... "contexts of use" and "interests and preferences"
- http://www.ht2009.org/ (deadline 2.02.) but: has high standards, so shouldn't be the only option
- K-CaP (deadline ???)
