The talk will introduce a geometric model and its investigation for contextual information retrieval. The geometric model leverages recent advances of vector space-based information retrieval. Some issues of the implementation of the model will also be discussed in the context of implicit feedback tailored for each task and personalized for each user. Besides the theoretical and methodological issues, the results of some experiments using rich interaction logs will also be described to show both the potential benefits of the proposed model and the value of incorporating multiple sources of interaction evidence when developing implicit feedback algorithms.