Places and place types, such as small town, play a fundamental role in how people organize knowledge about the world. Although places are commonly referenced in human communication, often they are not canonically defined and many of the properties people associate with them have proved difficult to operationalize. In information systems the representation of place knowledge has predominantly focused on either social science metrics or spatial properties such as location and spatial footprint. However, on the web there exist rich datasets containing unstructured natural language texts that are spatially referenced and which provide us with a window to the salient properties of places as people write about them.

This research investigates the operationalization of thematic place properties from collections of place descriptions and methods for reasoning with this knowledge in concert with other place information. This work is organized into three main aspects: place knowledge discovery, place reasoning, and the application of the previous two for geographic knowledge exploration. First, a procedure for extracting thematic place knowledge is presented using a novel combination of topic modeling and spatial statistics. The latent topical contents of documents are interpreted as observations at locations or regions and mapped into properties of places. This procedure is evaluated in terms of its utility for estimating the locations associated with non georeferenced texts. Second, a system is developed to automate reasoning about similar and analogous places, and to find real-world analogs to novel compositions of properties and places, such as "Denver with beaches." To systematize this reasoning, a novel observation-to-generalization place model that subsumes the digital gazetteer model is defined. This model differentiates between spatial observations of phenomena and place-based generalizations derived from these observations. It forms the basis for a suite of reasoning operations to find place analogs based on sets of invariant properties. In the third part, the engineering and architecture of Frankenplace, an online geographic knowledge exploration tool that utilizes these discovery and reasoning frameworks, is detailed. This application enables a user to explore the topics relevant to a place, find relevant regions based on a thematic keyword query, and search for analogous places.