An Ontology based Recommendation System

I. Zinnikus, C. Fraser, and K. Fischer (UK)

Keywords

Useradaptivity, user modeling,ontology, SAID.

Abstract

This paper presents the use of a hierarchical ontology representation as a basis for modeling user preferences in the SAID project for elderly and disabled persons. The nodes of the ontology tree are associated with either pre selected or dynamically found Web sites relevant for each topic of the user's interest (in this way Web surfing can be substituted by browsing through the ontology). In order to further reduce the complexity of the tree, we combine this approach with applying a probability based recommendation system which monitors the user's actions and adapts itself according to the observed behaviour. We show that the resulting ontology tree is compatible with Bayesian networks and discuss limitations and possible extensions. The user profile is extended by using a vector space model. Combining the probability approach with the vector space model, personal recommendations of interesting documents can be made.

Important Links:



Go Back