The objective of this proposal is to demonstrate a set of methods for automatically extracting metadata from diverse data sets to serve as a common vocabulary by which data can easily be queried, retrieved and combined for visualization in a geobrowser. We propose extracting keyword tags from both structured and unstructured data sets by applying natural language processing (NLP) to metadata and unstructured content. The extracted tags will be associated with each data set as supplementary metadata to assist with data discovery, categorization and spatial-temporal location. We combine manually-generated tags, based on domains of interest or specific decision support activities, with automatically generated tags from NLP, and to develop hierarchical clusters of the combined tags to serve as a common set of descriptors by which different data sets can be discovered and combined. If proven successful, our approach will be useful for the management and fusion of very large and diverse data sets not only for applied science and decision support, but also for emergency management and related security operations, for business intelligence, and for other application involving large quantities of diverse data, both structured and unstructured.