MetaCORE: Metadata automated Categorization and Optimized Relevance Exploration

To support net-centric warfare, the Air Force must automate the generation and maintenance of metadata about both new and legacy information products. Metadata will enable Warfighters to retrieve necessary information quickly enough for accelerated ops tempos from among the rapidly increasing number of accessible information products. The metadata must include standard, domain-defined, user-generated, and automatically-generated attributes. We will build the *Metadata automated Categorization and Optimized Relevance Exploration* (*MetaCORE*) application based on the powerful analysis and indexing methodology of /Probabilistic Latent Semantic Indexing/ (pLSI). pLSI both extracts indices from collections of documents and associates specific indices with individual documents. These indices are both explicit – i.e., contained in the document – and latent – i.e., /not/ contained in the document but associated with it because of its similarity to other documents. /pLSI can provide an innovative, effective, and efficient solution to the metadata problem by addressing its central challenges/. It is designed to determine indices in document collections and to associate appropriate indices with particular documents, thus providing automatically-generated attributes. We will also develop a simple tool by which users can prepare user-specified or domain-defined attributes. /pLSI can then use the same underlying technology to associate those additional kinds of metadata with individual documents/.