Robin Burke

home > research

Research

I have been working in the area of recommender systems since before the field had its name. One of my primary emphases has been hybrid recommendation -- the combination of multiple recommendation techniques and knowledge sources.

Differential Context Relaxation and Differential Context Weighting

One of the most difficult problems in context-aware recommendation is the identification and selection of contextual variables. DCR applies contextual variables in a differential way using different variables in different components of the recommendation algorithm. Optimization techniques are used to identify appropriate variables for each component.

Experience Discovery

This project is in its early phases, but we are working to recommend out-of-school time activities to students using a mix of collaborative, content and social network data.

Older projects

Secure Personalization

Can we make recommender systems secure against attackers injecting bias profiles? The answer is "not totally", but there is some interesting progress that can be made.

ARCH

User modeling for information access using ontologies. Project wiki.

FindMe systems

CBR, user modeling, and critique-based navigation applied to the problem of recommending products in e-commerce catalogs.

FAQ Finder

Natural language question answering from frequently-asked question files, combining statistical and knowledge-based techniques.

Donna/Salticus

Strategic knowledge applied to the problem of finding, storing, organizing and presenting information for competitive business intelligence.

Intelligent Tutoring Systems

My dissertation work was on intelligent tutoring systems, concentrating on the problem of storing, indexing and retrieving tutorial stories.