Much of the academic literature available on the Web has never been adequately catalogued. Consequently, even using large-scale search engines, much of it remains inaccessible to researchers as indexing on this scale lacks the necessary detail to cope with discipline dependent terminologies and ontologies. Metadata has become a popular means to provide such information within known domains. In this paper, we describe an approach to the automatic extraction of metadata from medical research papers. Medical research papers tend to have stereotypic prescribed sections, such as introduction, methods, and conclusions. The approach described uses context thesauri and the semantic structure of the documents to extract metadata based on these stereotypic sections.