User Model in a Box: Cross-System User Model Transfer for Resolving Cold Start Problems
Recommender systems face difficulty in cold-start scenarios where a new user has provided only few ratings. Improving cold-start performance is of great interest. At the same time, the growing number of adaptive systems makes it ever more likely that a new user in one system has already been a user in another system in related domains. To what extent can a user model built by one adaptive system help address a cold start problem in another system? We compare methods of cross-system user model transfer across two large real-life systems: we transfer user models built for information seeking of scientific articles in the SciNet exploratory search system, operating over tens of millions of articles, to perform cold-start recommendation of scientific talks in the CoMeT talk management system, operating over hundreds of talks. Our user study focuses on transfer of novel explicit open user models curated by the user during information seeking. Results show strong improvement in cold-start talk recommendation by transferring open user models, and also reveal why explicit open models work better in cross-domain context than traditional hidden implicit models.
KeywordsCross-system user modeling Recommender systems
Unable to display preview. Download preview PDF.
- 2.Cremonesi, P., Tripodi, A., Turrin, R.: Cross-domain recommender systems. In: 2011 IEEE 11th International Conference on Data Mining Workshops, pp. 496–503. IEEE, December 2011Google Scholar
- 3.Fernadez-Tobis, I., Cantador, I., Kaminskas, M., Ricci, F.: Cross-domain recommender systems: a survey of the state of the art. In: the 2nd Spanish Conference on Information RetrievalGoogle Scholar
- 4.Li, B., Yang, Q., Xue, X.: Can movies and books collaborate?: cross-domain collaborative filtering for sparsity reduction. In: Proceedings of the 21st International Jont Conference on Artifical Intelligence, IJCAI 2009, pp. 2052–2057. Morgan Kaufmann Publishers Inc, San Francisco (2009)Google Scholar
- 5.Loizou, A.: How to recommend music to film buffs: enabling the provision of recommendations from multiple domains. Thesis (2009)Google Scholar
- 6.Pan, W., Liu, N.N., Xiang, E.W., Yang, Q.: Transfer learning to predict missing ratings via heterogeneous user feedbacks. In: Proc. of the Twenty-Second International Joint Conference on Artificial Intelligence - Volume Three, IJCAI 2011, pp. 2318–2323. AAAI Press (2011)Google Scholar
- 7.Ruotsalo, T., Peltonen, J., Eugster, M., Głowacka, D., Konyushkova, K., Athukorala, K., Kosunen, I., Reijonen, A., Myllymäki, P., Jacucci, G., Kaski, S.: Directing exploratory search with interactive intent modeling. In: Proceedings of the 22nd ACM International Conference on Conference on Information & Knowledge Management, CIKM 2013, pp. 1759–1764. ACM, New York (2013)Google Scholar
- 8.Sahebi, S., Brusilovsky, P.: Cross-domain collaborative recommendation in a cold-start context: the impact of user profile size on the quality of recommendation. In: Carberry, S., Weibelzahl, S., Micarelli, A., Semeraro, G. (eds.) UMAP 2013. LNCS, vol. 7899, pp. 289–295. Springer, Heidelberg (2013) CrossRefGoogle Scholar
- 9.Schein, A.I., Popescul, A., Ungar, L.H., Pennock, D.M.: Methods and metrics for cold-start recommendations. In: Proceedings of the 25th Annual International ACM SIGIR Conference on Research and Development in Information Retrieval, SIGIR 2002, pp. 253–260. ACM, New York(2002)Google Scholar
- 10.Schwab, I., Pohl, W., Koychev, I.: Learning to recommend from positive evidence. In: Proceedings of the 5th International Conference on Intelligent User Interfaces, IUI 2000, pp. 241–247. ACM, New York (2000)Google Scholar
- 11.Shi, Y., Larson, M., Hanjalic, A.: Tags as bridges between domains: improving recommendation with tag-induced cross-domain collaborative filtering. In: Konstan, J.A., Conejo, R., Marzo, J.L., Oliver, N. (eds.) UMAP 2011. LNCS, vol. 6787, pp. 305–316. Springer, Heidelberg (2011) CrossRefGoogle Scholar