Ontology-Based Modeling and Recommendation Techniques for Adaptive
Hypermedia Systems (Teza de Doctorat)
-----------------------------------
          
ABSTRACT
-----------------------------------
The specific problem addressed by the thesis is the problem of
providing personalized recommendations of resources to the users
of an e-learning application, according their professional competences
and interests, with the support of semantic Web technologies,
and without requiring explicit feed-back or other intervention
from the user. This problem is situated at the confluence of
multiple computer science domains, whose technologies were harmonized
in order to structure the solution provided by the thesis: e-learning,
semantic Web, adaptive hypermedia systems.

In order to define thesis solution, some integrally similar approaches 
oriented to the same entire problem were not available. 
Instead, each step performed to the final solution was related to 
the existing approaches and standards which address the corresponding 
intermediary problem.

The main issues addressed for developing a recommender system concern 
the definition of user model and document model, the methods for these 
models development, the technique for establishing the recommendations 
suitable for each particular user, as well as the architecture to 
include the corresponding functionalities.

The main original aspects are:
(1) Using ontology as a binder between documents and users: the 
user profile is defined in terms of the same ontology as the 
document model.
(2) The combination of approaches belonging to different domains.
(3) User profile is structured on three levels, competences, 
interests and fingerprints, and is built on top of IEEE/PAPI. 
(4) Document model is built on top of IEEE/LOM, including three 
semantic relation defined. Different parts of the document 
were separately processed in order to generate semantic 
differentiated annotations. Also, a solution of workflow modeling 
for multimedia documents indexing was provided.
(5) The hybrid recommender solution adopts a new approach for 
using domain knowledge in the user and document knowledge to be 
applied in known recommender techniques. The method of integrating 
ontology-based domain knowledge into the recommender algorithms is 
original, as well as the idea of supervising the user conceptual 
navigation through the ontology used for materials and users. 
The recommender solution avoiding cold-start issue, and also 
does not demands user feed-back.
(6) Concerning the solution architectural issues, the defined 
models could be exported from a systems to another. 
Multiple architectures are possible for the recommender 
solution: service-oriented architecture adopted independently, 
or into a grid system, or inside a pervasive environment.

As conclusion, the thesis provides a semantic-based approach in 
the context of the new European higher education law to alleviate 
and facilitate the dissemination of the current overload of information 
and resources for teaching and learning in Computer Science.
          
SPEAKER(S)
-----------------------------------
Lect.dr. Mihaela BRUT
Universitatea Alexandru Ioan Cuza, Facultatea de Informatica
Iasi
Romania
-----------------------------------