The project is well suited for students who already have a background in applying machine learning.

On the SemanticDesktop? we talk a lot formalising and linking existing knowledge, i.e. crawling your emails and documents, and then linking the sender of the email with the author of documents.

If this new data-space that we create is so great, it should be well suited as input for machine-learning, and making your desktop learn should be easier than ever before.

Also, the Semantic Desktop offers many opportunities for learning techniques to make it better! For instance:

• Although Nepomuk goes a little way towards formalising your knowledge, there are clearly many more potential links between desktop resources that we do not find. Learn from existing relationships and propose new relations in the knowledgebase?
• Learn the assignment of tags to resources
  • Useful for tag suggestions!
• Learn to predict what resources a user may access next
  • We keep track of each resource the user selects, and many other things for UserWorkContext

I will not propose exactly what this project should try to achieve, if you have any ideas please come and discuss – I am happy to tutor almost anything :)

Tutor: GunnarGrimnes