We present the application of Latent Semantic Analysis (LSA) in combination with recommender systems, in order to enhance discovery in geoportals. As a basis for discovery, metadata of spatial data and services, as well as of non-spatial resources, such as documents and scientific papers, is created and registered in the catalogue of the geoportal (semi-)automatically. Links that are not inherent in the data itself are established based on the semantic similarity of its textual content using LSA. This leads to the transition from unstructured data to structured (metadata) information, serving as a basis for the generation of knowledge. The metadata information is integrated into a recommendation system that provides a ranked list showing (1) what other users viewed and (2) the related resources discovered by the LSA workflow as a result. Based on the assumptions that similar texts have something in common and that users are likely to be interested in what other users viewed, recommendations provide a broader, but also more precise, search result; on the one hand, the recommender engine considers additional information; on the other hand, it ranks resources based on the discovery experience of other users and the likeliness of the documents being related to each other.