Digital media in general and the Internet in particular play a ever growing role in the storage, dissemination and retrieval of information and knowledge. Semantic annotation of the provided information is necessary to support the user in retrieving and filtering requested information. Wide-spread tools like Google have shown both the power and the limitations of statistical methods based on correlations between terms within texts. The KEA system uses a different approach. Natural language processing techniques, taking advantage of characteristic linguistic structures defined by the language used in mathematical texts, are utilized in the automatic extraction of ontologies from mathematical texts. As a result, the system creates a base of the mathematical relations and concepts by mapping the extracted ontologies on the structure of a data base. To support the retrieval and further refinement of the data stored within this mathematical knowledge base, KEA implements a number of Web 2.0 technologies in a desktop style interface. The following article presents the technology and applications of the KEA system.