Recommender Systems

Week 1

Introduction to Recommender Systems

You can invest time in indexing information. Indexing can be done by comparing term frequency with in a document compared to the corpus.

History

• Invormation retrieval and filtering
• manual collaborative filtering
• automated collaborative filtering
• commercial era

Taxonomy of Recommender Systems

• Does sequence matter?
• Are we recommending new things? (an example of re-recommending, is Pandora)

A search engine is a recommender.

• What is the purpose of the recommendation? a sale, to inform

example: OWL - observed a word processor user and recommended shortcut keys based on their actions.

Personalization

• Generic (everyone gets the same recommendations)
• Demographic
• Ephemeral (matches current activity)
• Persistent (matches long-term interests)