KNOWLEDGE MANAGEMENT

AND DISCOVERY LAB

Outline of the course
Part I - Basics of recommenders:
    1. Goals, architecture and components of a recommender
    2. Recommenders in the context of Customer Relationship Management
Part II: Learning methods for the back-end of the recommendation engine
    1. Naive predictors
    2. Within-lab evaluation of a recommender
    3. Content-based recommenders
    4. CF-based recommenders
    5. MF-based recommenders
Part III: Trust building and explainations for the front-end of the recommendation engine
Part IV: Advanced topics
    1. Deriving ratings from opinions; learners for Sentiment Analysis & Opinion Target Extraction
    2. Further topics on machine learning for recommenders
We plan one exercise class on German (UE_DEU) and one on English (UE_ENG).


Timeplan of the course (tentative)
Meeting 1: Introduction to the course, administrative issues; Part I
Meeting 2: Part I
Meetings 3-8: Part II
    • Meeting 3: Naive methods (1) and principles of in-lab evaluation (2)
    • Meetings 4, 5: Content-based recommenders (3)
    • Meeting 6: CF-based recommenders (4)
    • Meetings 7, 8: MF-based recommenders (5)
Meeting 9: Part III
Meetings 10-13: Part IV
    • Meetings 10-12: Opinion mining (1), presentations of Option CHOICE (from Meeting 11 on)
    • Meeting 13: Live evaluation (2), presentations of Option CHOICE
Meeting 14: Q & A

Recommended background
Bachelor students: You should have already attended the compulsory Computer Science courses of the early semesters.
All students: Having attended data mining or machine learning or a course on artificial intelligence is of advantage, although you may catch up with home reading. It is strongly advisable that you refresh your secondary school background in statistics.

Goal of this course is to make you familiar with recommenders. You will learn what requirements are placed to a recommender by the business operating it and by the users interacting with it, and you will become proficient in methods used to meet these requirements.
A recommender has a front-end service responsible for the interaction, and a back-end machine learning core that derives the recommendations to be presented to the users and learns from the users' behavior. Most part of the course is on the back-end core: you will become familiar with basic, heuristic methods and with advanced methods that model the recommendation task as an optimization problem and deliver solutions for it.
A recommender exploits information on the items to be recommended and information on the users to be served. You will see methods that extract such information from data. The course's main emphasis will be on web data, including opinionated texts from social fora.

Book of the course:
F. Ricci, L. Rokach and B. Shapira (2011) Introduction to Recommenders Systems Handbook, SPRINGER
              
Core papers on the underpinnings of specific course topics:
    • J.L. Herlocker, J.A. Konstan, L.G. Terveen and J.T. Riedl (2004) Evaluating Collaborative Filtering Recommender Systems, ACM Trans. On Information Systems, Vol. 22, 5-53, 2004 (on: Evaluation)
    • M. Hu and Bing Liu (2004) Mining and summarizing customer reviews, Proc. of ACM Int. Conf on Knowledge Discovery from Data (KDD'04), Test-of-time Award at KDD'15 (on: Opinion Mining)
    • G. Takács , I. Pilászy, B. Németh and D. Tikk (2009) Scalable Collaborative Filtering Approaches for Large Recommender Systems, Journal of Machine Learning Research 10, 623-656 (on: CF & MF)
Further papers are cited in the materials of the course.
The papers for the RECSYS-6ects homework assignment (options CHOICE and DEFAULT) will be announced during the Part IV of the course.

None.

RECSYS is for bachelor students in high semesters (5 ECTS) and master students (6 ECTS) of:
    • Bachelor INF, INGINF, CV and Bachelor WIF (areas: Gestalten & Anwenden)
    • Master INF, INGINF, DigiEng and Master WIF (area: Wirtschaftsinformatik)
    • Data Science Master DKE (areas: Methods I and Applications)
Under Exam you find how you acquire the 5 ECTS, resp. 6 ECTS.