Exposing, Formalizing, and Reasoning over the Latent Semantics in Multimodal Domains

We aim to expose latent semantic information in folksonomies by mapping their tags through lexicons to ontologies and by applying other strategies, tools, and techniques to achieve high relevance and fidelity. We will analyze the problem space, select an appropriate domain and tools, and provide a detailed design for a system to provide the ability to reason over this latent semantic information. Our work will include the use of linguistic and other tools to permit rapid, accurate, and reliable mappings from “folksonomic” structures with weak or nonexistent semantics to formal ontologies, where high-level and sound reasoning will then be applicable to information previously opaque to such processing. We believe that an accurate and reliable methodology for mapping implicit semantics into explicit semantics is a critical factor in fulfilling the promise of the Semantic Web, , and that successful completion of our work plan will result in novel and powerful tools with significant value and relevance across a broad spectrum of applications, covering both military/intel and civilian/commercial domains and applications.