Personalized Click Model through Collaborative Filtering
Authors: Si Shen, Botao Hu, Weizhu Chen, and Qiang Yang
In Proceedings of the 5th ACM International Conference on Web Search and Data Mining (WSDM), Seattle, WA, USA, 2012.
Latest advances in click model proclaimed itself a promising approach to the interpretation of search click data. Existing models can well characterize document (i.e. url snippet) position bias and major click behavior. Yet, current development depicts user search actions in a general setting by assuming that all users act in a same way, regardless of the fact that one, with a particular interest, is more likely to click a link than the rest of the population. It is in light of this that we put forward a novel personalized click model to describe user-oriented click preferences, which applies and extends matrix / tensor factorization from the view of collaborative filtering to connect users, queries and documents together. Our model serves as a generalized personification framework that can be incorporated to the previously proposed click models and perhaps to their future expansions. Despite the sparsity of search click data, we present its advantage over the best click models in Web Search literature after a large-scale experiment on a real dataset. A delightful bonus is the model’s ability to get insights of queries and documents through latent feature vectors, and hence to handle rare and even new query-document pairs, that preceding click models could only take an average value.