The UC Berkeley Digital Library Project is building work-centered digital information services, creating  tools for user-oriented web access to large, distributed collections of diverse data.  As a research corpus, we are assembling a testbed on the environment of the State of California containing: 240,000 pages of OCR'd text (55 Gb), 60,000 color slides of native plants and other environmental features (330 Gb), GIS layers, satellite imagery and ortho-photos (35 Gb), and several RDBs including a Botanical Database of 169,000 plant occurrences.  Our testbed has been made available on-line as work evolved; and we have a user community of scientists, land planners, and government officials responsible for environmental protection and natural disaster planning. Our strategy is for all data to be accessible via DBMS with enhanced functionality, and with transparent delivery via WWW interfaces and "Thick" Java clients.

Accomplishments include: development of the prototype for "Multivalent Documents" (MVD), a  GIS Viewer , a Natural Language Processing interface: Tilebars, and advances in Vision-based image retrieval.  Currently we are working on remote, distributed annotation, on TilePix satellite image resampling and visualization , and access to remote heterogeneous GIS datastores with on-the-fly re-projection and software translation.