Recommender Systems

Question 1

Diff between Information retrieval and Information filtering

Answer 1

Retrieval serves varying user needs from static DB (search).

Filtering serves varying incoming (streaming) content for static user needs.

Question 2

Collaborative filtering

Answer 2

‘GroupLens’ got reviews on news articles from users, and recommended what other people of similar taste liked.
Manual CF: Users in groups FW jokes/articles to each other

‘MovieLens’ exists to recommend based on your taste

Question 3

Dimensions of reco?

Answer 3

1. Content: News, products, bundles, people, sequences (playlists)
2. Purpose of reco: Buy, consume, learn (educate people on what they are missing on a product for e.g.)
3. Context: What people doing at time of reco
4. Personalization level: Demographic, Ephemeral, Long-term
5. Privacy and trustworthiness: Do I trust business rules from the recommender say if it is hiding low markup products
6. Interface: Input and Output
7. Algos: Non -personalized, Content based filtering, Collab filtering, interviews/hybrid