Using natural language processing, the document describes an attempt to automatically identify topics within a corpus of 812 papers related to ecosystem services. Latent Dirichlet Allocation was used to extract 80 topics from the text. The top words for each topic were analyzed and many closely matched the categories of ecosystem services established by the Millennium Ecosystem Assessment. However, some topics did not cleanly fit into the MEA framework, highlighting challenges in applying computerized topic modeling across disciplines.