A system that takes a free-form text query for places and tries all possible interpretations to find a match based (somehow) on a history of places for that user (history in the general term, the way I read it, it can be places the user has been, places the user has searched, places whose shape look like the user's face, etc) and a global geospatial database.

So, a fancy parser, a per-user list of places of potential interest and a Solr collection with geospatial data.

P.S. the irony is that at no point in that overly verbose spec, is there anything that in any way tells you how to build that system. It's like patenting a device for enabling one to force a resistance against achieving minimum possible potential energy in a gravity field upon a mass of fluid, i.e. the concept of a bucket (or a glass, or the act of cupping your palms). And while that is surely only the preface of the patent, I seriously doubt the body is any more illuminating. Maybe while completely abolishing software patents remains impossible, we should try to enact reform that forces them to be specific enough that just by reading the patent one would be able to write a program that performs the actions described therein. Even if it's just "take a solr server, configure it like this, take this particular natural language parser, generate the possible forms of the query like so...". At least then they would be useful in a decade or so.