Geographically referenced user generated content provides us with an opportunity to, for the first time, gather perspectives on place over large areas by exploring how very many people describe information. We present a framework for analysing large collections of user generated content. This involves classification of descriptive terms attached by users to photographs into facets of elements, qualities, and activities. We apply this framework to two contrasting photographic archives — Flickr and Geograph, representing weakly and strongly moderated content respectively. We propose a method for removing user–generated bias from such collections though the user of term profiles that can assess the effect of the most and least prolific contributors to a collection. Analysis and visualization of co–occurrence between terms suggests clear differences in the description of place between the two collections, both in terms of the facets used and their geographical footprints. This is attributed to the role of moderation/editorialising of content; to the role tags and free–text have on descriptive behaviour and to the geographic footprint of content supplied by the two collections.