Folksonomies and Ontologies in Authoring of Adaptive Hypermedia Fawaz Ghali, Mike Sharp, and Alexandra Cristea
Overview The Problem The Solution Advantages Methodology Results Findings
The Problem The ontologies lacks of flexibility, scalability. The folksonomies consists of unstructured data and carries no semantics, leading to questions of accuracy and reusability.
State of the art Modified from: www.mindingtheplanet.net
Solution Merging folksonomies from the Social Web with ontologies from the Semantic Web.
Advantages Creating semantic relations between tags of folksonomy. Augmenting the authoring of adaptive hypermedia
Methodology
Filtering Google API Spell Checker
Grouping
Mapping Swoogle (SW Search engine) Jena (SW Framework)
Experiment Flickr Tag Cloud Flickr API Dataset 11,138 tags
Conditions Each tag has to occur at least ten times to be a part of a group Every two groups that share more than five tags must be combined into one group, to avoid redundancy
Results {Food Fruit} {Desert food:CheeseNutsDessert} {vin: property colour = Red} {Meat RedMeat NonSpicyRedMeat} {RedMeat NonSpicyRedMeat}
Results Hierarchical cluster analysis for the food group
Findings Decreasing the cost of authoring semantic content. Tags are context-specific Tags can be system-specific
Findings All tags are connected to each other Not all social web tag-groups are covered by ontologies Several groups share the same subset of tags
Findings The most used ontology element was “instance”. Merging knowledge from multiple ontologies could provide a much richer perspective It is possible to have tags in a group which are not mapped onto same ontology.
Using the Authored Hierarchical Structure Enriching the authoring environments Enrich the adaptive strategies IF (concept.tag == UM.tag) then concept.show = TRUE Suggestions and/or corrections
Questions?