The Net of the Twenty-First Century must permit users to directly solve their information problems. Hypermedia browsing has now become widespread and search facilities are beginning to appear. Users are now building information repositories on a grand scale. This will soon lead to a global information space consisting of a billion repositories.

(See Evolution of the Net.) What will this future world be like? How will we locate and correlate information in such a vast space?

The Interspace Research Project is developing a prototype environment for semantic indexing of multimedia information in a testbed of real collections. The semantic indexing relies on statistical clustering for concepts and categories. Interactive navigation based on semantic indexing enables information retrieval at a deeper level than previously possible for large, diverse collections.We are in the process of developing algorithms for automatically extracting concepts and computing Concept Spaces, Category Maps, and performing Concept Assignment. Our collections include engineering literature, map images, and medical literature. The Interspace Prototype will thus enable scalable, interactive semantic interoperability across subject domain, media type, and collection size.

.....