Medial node-based centrality measures consider paths that start or end at a given node.

Monolithic hybridization designs always use late fusion as aggregation approach.

In a meta-level hybrid the first recommender builds a model that is passed to the second recommender, while a cascade hybrid introduces additional items in the ultimate recommender of its pipeline.

Borda rank aggregation is a late-fusion method that can be used in parallelized hybridization designs.

Reasons to use hybrid recommender systems include: (1) their capability to reflect different (complementary) characteristics of the items in the target domain, (2) their potential to improve accuracy and beyond-accuracy measures. and (3) their potential to mitigate cold start situations.

Below you find some examples for different categories of contextual data that can be used in a context-aware (or hybrid) recommender system. Please connect the examples with the six categories discussed in lecture.

Active user awareness requires the user of a context-aware recommender system to trigger a context-aware update of the recommendation list.