Courses in summer term 2004 / Seminar on Recommender Systems:

Time:	Wednesday 14-16
Location:	SR 01-018, Geb. 101
Begin:	21.4.2004

Recommender Systems are an intelligent access technology to large information systems as online catalogs in e-commerce or digital libraries and have been identified as one of the key technologies for e-commcerce. Recommender systems try to recommend users items that are of specific interest for them, based on user profiles of an online community build from explicit ratings of products or implicit usage information. Recommender systems may be as simple and ubiquitous as Amazons "who bought this, also bought that" crosslinks, and they may be rather complex knowledge and data driven systems aiming at modelling human counselors. Thus, recommender systems are the probably most advanced technology for personalization, drawing input from disciplines as heterogenous as e-commerce, online information systems, dynamic web technologies, data mining, information retrieval, articifical intelligence, user modelling, and human computer interaction.

The seminar gives a broad overview of different technologies and methods used for modeling, building, and deploying recommender systems.

Talks can be given in English or German.

Topics (M = methodological, T = technological focus):

0. -- Introduction --
1. (M) Collaborative filtering.
2. (M) Content-based filtering and hybrid filtering methods.
3. (M) Markov decision processes for modeling recommender systems.
4. (M) profiling and implicit feedback
5. (T) user models and ontologies.
6. (M) recommender systems and network theory (hubs and authorities).
7. (M) recommender systems and social choice theory.
8. (M) conditional preference nets (CP-nets).
9. (T) Adaptive web sites, assisted browsing, and web caching and prefetching.
10. (T) electronic program guides in digital tv.
11. (T) Recommender systems in e-commerce.
12. (T) Recommender systems in digital libraries and e-learning.

For more information on the topics see readings. You can register for a topic by email from now.