1. Dataset-driven research to improve TEL recommender systems Katrien Verbert, HendrikDrachsler, Nikos Manouselis, Martin Wolpers, RiinaVuorikari and Erik Duval

2. What is dataTEL? dataTEL is a Theme Team funded by the STELLAR network of excellence. It addresses 2 STELLAR Grand Challenges Connecting Learner Contextualization

3. dataTEL::Objective Five core questions: How can data sets be shared according to privacy and legal protection rights? How to develop a respective policy to use and share data sets? How to pre-process data sets to make them suitable for other researchers? How to define common evaluation criteria for TEL recommender systems? How to develop overview methods to monitor the performance of TEL recommender systems on data sets? Standardize research on recommender systems in TEL

4. Free the data B Tom Raftery http://www.flickr.com/photos/traftery/4773457853/sizes/l

5. Why? By Tom Raftery http://www.flickr.com/photos/traftery/4773457853/sizes/l

6. Because we will get new insights By Tom Raftery http://www.flickr.com/photos/traftery/4773457853/sizes/l

8. dataTEL challenge & dataTEL cafe event a call for TEL datasets eight datasets submitted http://bit.ly/ieqmWW

9. http://dev.mendeley.com/datachallenge/

11. Collaborative filtering Users who bought the same product also bought product B and C

12. User-based CF A Sam high correlation B Ian C Neil

13. Item-based CF A Sam B high correlation Ian C Neil

14. similarity measures Cosine similarity Pearson correlation Tanimoto or extended Jaccard coefficient

15. evaluation metrics Accuracy: precision, recall, F1 Predictive accuracy: MAE, RMSE Coverage

16. experiments Collaborative filtering based on ratings Collaborative filtering based on implicit relevance data

17. similarity measures MAE of item-based collaborative filtering based on different similarity metrics

18. algorithms MAE of user-based, item-based and slope-one collaborative filtering Nikos Manouselis, Riina Vuorikari, and Frans Van Assche. Simulated analysis of MAUT collaborative filtering for learning object recommendation (SIRTEL07)

19. implicit relevance data F1 of user-based collaborative filtering with increasing number of neighbors

20. data dimensions

21. CEN WS-LT Social Data standardized representation of both explicit and implicit relevance data http://bit.ly/ho1MbC

22. Data set framework to monitor performance 22

23. evaluation criteria 1. Reaction of learner 2. Learning improved 3. Behaviour 4. Results 1. Accuracy 2. Coverage 3. Precision 1. Effectiveness of learning 2. Efficiency of learning 3. Drop out rate 4. Satisfaction Kirkpatrick model by Manouselis et al. 2010 Combine approach by Drachsler et al. 2008

24. So what about you… Do you have data that can be shared for research? Do you want to be involved in dataTEL research? datafortel@gmail.com

25. dataTEL challenge at I-KNOW 2011 11th International Conf. on Knowledge Management and Knowledge Technologies 7–9 September 2011, Messe Congress Graz, Austria

26. Many thanks for your attention! Slides are available at: http://www.slideshare.net/kverbert Email: katrien.verbert@cs.kuleuven.be Skype: katrien.verbert Twitter: katrien_v