We present a method to geometrize massive data sets from search engines query logs. For this purpose, a macrodynamic-like quantitative model of the Information Retrieval (IR) process is developed, whose paradigm is inspired by basic constructions of Einstein's general relativity theory in which all IR objects are uniformly placed in a common Room. The Room has a structure similar to Einsteinian spacetime, namely that of a smooth manifold. Documents and queries are treated as matter objects and sources of material fields. Relevance, the central notion of IR, becomes a dynamical issue controlled by both gravitation (or, more precisely, as the motion in a curved spacetime) and forces originating from the interactions of matter fields. The spatio-temporal description ascribes dynamics to any document or query, thus providing a uniform description for documents of both initially static and dynamical nature. Within the IR context, the techniques presented are based on two ideas. The first is the placement of all objects participating in IR into a common continuous space. The second idea is the `objectivization' of the IR process; instead of expressing users' wishes, we consider the overall IR as an objective physical process, representing the IR process in terms of motion in a given external-fields configuration. Various semantic environments are treated as various IR universes.